Resource return processing method and device, electronic equipment, storage medium and program product

Abstract

This application provides a method, apparatus, electronic device, storage medium, and program product for resource return processing. The method includes: in response to a resource return request, acquiring a historical transaction data set and available resource quantity of a target resource account; determining a necessity score based on the attribute characteristics of each transaction data unit, and mapping multiple transaction data units to multiple hierarchical categories based on the necessity score; determining the proportion of survival-type transactions based on transaction data units belonging to the survival-type category, and determining the resource pressure level of the target resource account based on the proportion of survival-type transactions; determining a comprehensive score for each transaction data unit based on the necessity score and transaction attributes; and generating a combined resource return strategy including segmented return instructions based on the resource pressure level and the comprehensive score of each transaction data unit when the available resource quantity is less than the total resource quantity of the historical transaction data set. This application improves the rationality of resource return strategies.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to a method, apparatus, electronic device, storage medium, and program product for resource return processing. Background Technology

[0002] With the rapid development of electronic finance, more and more users are using credit payments. When repaying debts, users face a variety of repayment methods, each with different debt requirements. Related technologies typically determine repayment strategies based on fixed rules, calculating the minimum repayment amount and installment fees based on historical transaction data and credit rating, and providing users with standardized options and recommending fixed repayment methods. Summary of the Invention

[0003] This application provides a method, apparatus, electronic device, storage medium, and program product for resource return processing, which can improve the rationality of resource return strategies.

[0004] The technical solution of this application embodiment is implemented as follows: This application provides a method for processing resource return, the method comprising: In response to a resource return request for a target resource account, the system obtains a set of historical transaction data and the amount of available resources for the target resource account, wherein the set of historical transaction data includes multiple transaction data units. Based on the attribute characteristics of each transaction data unit, a necessity score for each transaction data unit is determined, and the multiple transaction data units are mapped to multiple hierarchical categories based on the necessity score, wherein the hierarchical categories include at least a survival category. Based on the transaction data units belonging to the survival category, the proportion of survival transactions is determined, and the resource pressure level of the target resource account is determined based on the proportion of survival transactions. Based on the necessity score and transaction attributes of each transaction data unit, a comprehensive score for each transaction data unit is determined. If the available resources are less than the total resources of the historical transaction data set, a combined resource return strategy containing segmented return instructions is generated based on the resource pressure level and the comprehensive score of each transaction data unit.

[0005] This application provides a resource return processing apparatus, including: The data acquisition module is used to respond to a resource return request for a target resource account by acquiring the historical transaction data set and available resource quantity of the target resource account, wherein the historical transaction data set includes multiple transaction data units; The hierarchical category determination module is used to determine the necessity score of each transaction data unit based on the attribute characteristics of each transaction data unit, and to map the multiple transaction data units to multiple hierarchical categories based on the necessity score, wherein the hierarchical categories include at least the survival category. The resource pressure level determination module is used to determine the proportion of survival-type transactions based on the transaction data units belonging to the survival-type category, and to determine the resource pressure level of the target resource account based on the proportion of survival-type transactions. The comprehensive score determination module is used to determine the comprehensive score of each transaction data unit based on the necessity score and transaction attributes of each transaction data unit. The strategy generation module is used to generate a combined resource return strategy containing segmented return instructions based on the resource pressure level and the comprehensive score of each of the transaction data units, when the available resource amount is less than the total resource amount of the historical transaction data set.

[0006] This application provides an electronic device, the electronic device comprising: Memory is used to store executable instructions or computer programs. The processor, when executing computer-executable instructions or computer programs stored in the memory, implements the resource return processing method provided in the embodiments of this application.

[0007] This application provides a computer-readable storage medium storing a computer program or computer-executable instructions, which, when executed by a processor, implements the resource return processing method provided in this application.

[0008] This application provides a computer program product, including a computer program or computer executable instructions. When the computer program or computer executable instructions are executed by a processor, they implement the resource return processing method provided in this application.

[0009] The embodiments of this application have the following beneficial effects: By applying the embodiments of this application, in response to resource return requests for the target resource account, historical transaction data sets and available resource quantities are acquired and analyzed, providing detailed data support for subsequent strategy formulation. By analyzing the attribute characteristics of each transaction data unit and calculating necessity scores, the data is mapped to hierarchical categories, including survival categories, achieving accurate identification and hierarchical management of the essential attributes of historical transaction behaviors, distinguishing the transaction data units necessary to maintain the basic survival of the account entity. The proportion of survival transactions is calculated using transaction data units belonging to the survival category, thereby determining the resource pressure level of the target resource account, transforming the abstract account status into quantifiable technical indicators, and objectively representing the target resource account's current available resource quantity. The system assesses the pressure and risk level under constraints; it combines necessity scores and transaction attributes to calculate the comprehensive score of each transaction data unit, providing a fine-grained quantitative basis for evaluating the priority and weight of each resource return; in a specific resource shortage scenario where the available resources are less than the total resources of the historical transaction data set, it intelligently generates a combined resource return strategy including segmented return instructions based on the resource pressure level and the comprehensive score of each transaction data unit, achieving dynamic adaptation of the resource return scheme. Through the synergistic effect of the combined resource return strategy and segmented return instructions, the system ensures the survival capability of the target resource account while orderly completing the resource return task, achieving an effective balance between resource utilization efficiency and return risk control, and improving the rationality of the resource return strategy. Attached Figure Description

[0010] Figure 1 This is a schematic diagram illustrating the application mode of the resource return processing method provided in the embodiments of this application; Figure 2 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application; Figure 3 This is a first flowchart illustrating the resource return processing method provided in the embodiments of this application; Figure 4 This is a schematic diagram of the second process of the resource return processing method provided in the embodiments of this application; Figure 5 This is a schematic diagram of the third process of the resource return processing method provided in the embodiments of this application; Figure 6 This is a schematic diagram of the fourth process of the resource return processing method provided in the embodiments of this application; Figure 7 This is a schematic diagram of the fifth process of the resource return processing method provided in the embodiments of this application; Figure 8 This is a schematic diagram of the sixth process of the resource return processing method provided in the embodiments of this application; Figure 9This is a schematic diagram of the seventh process of the resource return processing method provided in the embodiments of this application; Figure 10 This is the eighth flowchart of the resource return processing method provided in the embodiments of this application; Figure 11 This is a ninth flowchart illustrating the resource return processing method provided in this application embodiment; Figure 12 This is a schematic diagram of the tenth process of the resource return processing method provided in the embodiments of this application; Figure 13 This is an operational schematic diagram of the resource return processing method provided in the embodiments of this application; Figure 14 This is a schematic diagram of the eleventh step of the resource return processing method provided in the embodiments of this application; Figure 15 This is a first example diagram generated by the strategy provided in the embodiments of this application; Figure 16 This is a second example diagram generated by the strategy provided in the embodiments of this application.

[0011] It should be noted that the terms "first" and "second" mentioned above are only used to distinguish between different options and do not represent the degree of superiority or inferiority of the options or their priority in the implementation process. Detailed Implementation

[0012] To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings. The described embodiments should not be regarded as limitations on this application. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0013] In the following description, references are made to “some embodiments,” which describe a subset of all possible embodiments. However, it is understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0014] In the following description, the terms "first, second, third" are used merely to distinguish similar objects and do not represent a specific ordering of objects. It is understood that "first, second, third" may be interchanged in a specific order or sequence where permitted, so that the embodiments of this application described herein can be implemented in an order other than that illustrated or described herein.

[0015] In the embodiments of this application, the terms "module" or "unit" refer to a computer program or part of a computer program that has a predetermined function and works with other related parts to achieve a predetermined goal, and can be implemented wholly or partially using software, hardware (such as processing circuitry or memory), or a combination thereof. Similarly, a processor (or multiple processors or memory) can be used to implement one or more modules or units. Furthermore, each module or unit can be part of an overall module or unit that includes the functionality of that module or unit.

[0016] Unless otherwise defined, all technical and scientific terms used in the embodiments of this application have the same meaning as commonly understood by one of ordinary skill in the art. The terminology used in the embodiments of this application is for the purpose of describing the embodiments of this application only and is not intended to limit this application.

[0017] In this application embodiment, the relevant data collection and processing (e.g., the historical transaction data set and available resource quantity of the target resource account) should be strictly in accordance with the requirements of relevant laws and regulations when applied in practice. The informed consent or separate consent of the personal information subject should be obtained, and subsequent data use and processing should be carried out within the scope of laws and regulations and the authorization of the personal information subject.

[0018] Before providing a further detailed description of the embodiments of this application, the nouns and terms involved in the embodiments of this application will be explained, and the nouns and terms involved in the embodiments of this application shall be interpreted as follows.

[0019] 1) Transaction data unit: refers to the digital information record corresponding to each independent resource transfer behavior in the historical transaction records of the target resource account, including attributes such as transaction amount, transaction time, and merchant information.

[0020] 2) Combined resource repayment strategy: refers to an execution plan scheme that includes one or more repayment methods (such as minimum repayment, full repayment, installment repayment, etc.) generated by algorithmic comprehensive evaluation based on the current available resource status and repayment ability of the target resource account.

[0021] In the field of resource repayment processing, related technologies typically treat historical transaction data sets as a whole, lacking the ability to perform fine-grained analysis of the inherent attributes and importance of each transaction data unit within the set. This results in an inaccurate assessment of the true resource pressure on the target resource account, failing to reflect the implicit shortage of available resources caused by an excessively high proportion of essential transactions. Therefore, when available resources are insufficient to cover the total resource volume, only fixed and standardized repayment strategies can be provided, such as full repayment or minimum repayment, without prioritizing based on the urgency and necessity of each transaction data unit. This inherent strategy cannot achieve dynamic and differentiated resource allocation, easily leading to unreasonable resource allocation or increasing the potential risk of users defaulting.

[0022] This application provides a method, apparatus, electronic device, storage medium, and program product for resource return processing, which can improve the rationality of resource return strategies.

[0023] The following describes exemplary applications of the electronic devices provided in the embodiments of this application. These devices can be implemented as various types of terminals such as laptops, tablets, desktop computers, set-top boxes, smartphones, smart speakers, smartwatches, smart TVs, and in-vehicle terminals, or as servers. Exemplary applications when the device is implemented as a terminal or server will be described below.

[0024] See Figure 1 , Figure 1 This is a schematic diagram illustrating the application mode of the resource return processing method provided in this application embodiment. It is an example to support a resource return processing application. Figure 1 The system involves server 200, network 300, and terminal device 400. Terminal device 400 is connected to server 200 through network 300. Network 300 can be a wide area network, a local area network, or a combination of both.

[0025] In some embodiments, when the server 200 responds to a resource return request triggered by a user through the terminal device 400, it parses the historical transaction data set of the target resource account using the resource return processing method provided in this application embodiment. Each transaction data unit in the historical transaction data set is evaluated through attribute feature analysis to obtain a necessity score for each transaction data unit. Then, based on the necessity score, multiple transaction data units are mapped to multiple hierarchical categories, and transaction data units belonging to the survival category are identified to determine the survival transaction ratio. Next, the resource pressure level of the target resource account is determined based on the survival transaction ratio, and each transaction data unit is weighted based on the necessity score and transaction attributes to obtain a comprehensive score for each transaction data unit. Finally, when the available resource amount is less than the total resource amount of the historical transaction data set, a combined resource return strategy containing segmented return instructions is generated based on the resource pressure level and the comprehensive score of each transaction data unit and returned to the terminal device 400 to complete the rational allocation of resources for the target resource account.

[0026] See Figure 2 , Figure 2 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application. Figure 2 The server 200 shown includes at least one processor 410, memory 450, and at least one network interface 420. The various components of server 200 are coupled together via a bus system 440. It is understood that the bus system 440 is used to implement communication between these components. In addition to a data bus, the bus system 440 also includes a power bus, a control bus, and a status signal bus. However, for clarity, ... Figure 2 The general labeled all buses as Bus System 440.

[0027] Processor 410 can be an integrated circuit chip with signal processing capabilities, such as a general-purpose processor, a digital signal processor (DSP), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor can be a microprocessor or any conventional processor, etc.

[0028] The memory 450 may be removable, non-removable, or a combination thereof. Exemplary hardware devices include solid-state storage, hard disk drives, optical disk drives, etc. The memory 450 may optionally include one or more storage devices physically located away from the processor 410.

[0029] The memory 450 may include volatile memory or non-volatile memory, or both. The non-volatile memory may be read-only memory (ROM), and the volatile memory may be random access memory (RAM). The memory 450 described in this application embodiment is intended to include any suitable type of memory.

[0030] In some embodiments, memory 450 is capable of storing data to support various operations, examples of which include programs, modules, and data structures or subsets or supersets thereof, as illustrated below.

[0031] Operating system 451 includes system programs for handling various basic system services and performing hardware-related tasks, such as the framework layer, core library layer, driver layer, etc., for implementing various basic business functions and handling hardware-based tasks; The network communication module 452 is used to reach other electronic devices via one or more (wired or wireless) network interfaces 420, exemplary network interfaces 420 including Bluetooth, WiFi, and Universal Serial Bus (USB).

[0032] In some embodiments, the apparatus provided in this application can be implemented in software. Figure 2 A resource return processing device 455 stored in memory 450 is shown. This device can be software in the form of programs and plug-ins, and includes the following software modules: a data acquisition module 4551, a hierarchy category determination module 4552, a resource pressure level determination module 4553, a comprehensive score determination module 4554, and a strategy generation module 4555. These modules are logically connected and can therefore be arbitrarily combined or further divided according to their implemented functions. The functions of each module will be described below.

[0033] The resource return processing method provided in this application will be described by referring to the exemplary application and implementation of the electronic device provided in the embodiments of this application.

[0034] The resource return processing method provided in the embodiments of this application will be described below. As mentioned above, the electronic device implementing the resource return processing method of the embodiments of this application can be a terminal, a server, or a combination of both. Therefore, the executing entity of each step will not be described again below.

[0035] See Figure 3 , Figure 3 This is a first flowchart illustrating the resource return processing method provided in this application embodiment, which will be combined with... Figure 3 The steps shown are explained below. Figure 3 The implementing body is Figure 1 Server 200 in the middle.

[0036] In step 101, in response to a resource return request for the target resource account, the historical transaction data set and available resource quantity of the target resource account are obtained.

[0037] Here, the historical transaction data set includes multiple transaction data units. A resource return request is a state change instruction used to trigger the transfer of idle resources from the associated storage pool to the target resource account, in order to offset or balance the current resource occupancy status of the target resource account caused by historical operations. The purpose of a resource return request is to reduce the resource occupancy value of the target resource account to the expected threshold or to zero, thereby restoring the resource circulation capacity of the account.

[0038] In practice, the server is configured with a response receiving interface for the front-end interactive interface. When a specific graphical control (e.g., a resource return button) on the terminal device is triggered, it receives a signal containing the unique identifier of the target resource account and the request timestamp. Alternatively, the server runs a periodic scheduling task that, by traversing the account status table in the system database, filters out accounts whose resource settlement cycle has reached the critical point and whose current resource usage value is greater than zero, and automatically generates resource return requests for these filtered accounts.

[0039] Upon receiving a resource return request for a target resource account, a structured query instruction is constructed based on the unique identifier of the target resource account. This instruction retrieves all resource outflow records within a defined time window from the resource flow record database. The defined time window is defined as the period between the time node of the last resource settlement completion and the time node of the current request generation. Each retrieved resource outflow record is instantiated as a transaction data unit, and all instantiated units are combined to form a historical transaction data set. The specific storage format of the transaction data unit can be unstructured data stored in JSON or XML text format, or it can be a high-dimensional feature vector stored in a tensor database after preprocessing; this application does not impose any restrictions.

[0040] The target resource account is a virtual account entity registered in the resource management system that has the function of recording resource inflows and outflows. The historical transaction data set is a digital summary of all resource outflows or resource usage behaviors generated by the target resource account within a specific time period. It consists of multiple independent transaction data units, each of which encapsulates the detailed attributes of a single resource interaction operation through a specific data structure. Available resources are the amount of resources in the source resource pool that are bound to the target resource account, are not currently frozen, and have the authority to transfer them to the target resource account for replenishment operations.

[0041] Based on the pre-established association mapping relationship of the target resource accounts, the source resource pool account, which serves as the source of resource supply, is identified. By calling the status read interface of the core resource management system, the current active and unlocked resource values ​​in the source resource pool account are queried in real time. These resource values ​​are then determined as the available resource quantity in this processing flow, used to subsequently determine whether the resource return requirements are met.

[0042] Furthermore, in some high-concurrency scenarios that process massive amounts of data, the process of obtaining historical transaction data sets can be carried out based on a time-sharded columnar storage database, or by accessing hot data snapshots in an in-memory database instead of direct disk scanning, in order to reduce input and output latency. At the same time, a distributed lock mechanism or pre-emption mechanism can be introduced into the logic for obtaining available resources. That is, while querying values, a corresponding number of resources are temporarily locked to prevent these resources from being preempted by other parallel processes during subsequent calculations, thereby ensuring data consistency.

[0043] By employing the above method, the full details of historical resource usage and the current available resource inventory are obtained simultaneously in a single response step, constructing a complete data context environment. This provides an accurate and real-time data benchmark for subsequent resource matching and transfer calculations, effectively avoiding resource return calculation deviations caused by missing data dimensions or asynchronous query delays. At the same time, through standardized transaction data unit encapsulation, the compatibility and parsing efficiency of unified processing of resource interaction records of different types and sources are improved.

[0044] In step 102, based on the attribute characteristics of each transaction data unit, the necessity score of each transaction data unit is determined, and multiple transaction data units are mapped to multiple hierarchical categories based on the necessity score.

[0045] In some embodiments, see Figure 4 , Figure 4 This is a schematic diagram of the second process of the resource return processing method provided in the embodiments of this application; Figure 3 Step 102, "Determining the necessity score of each transaction data unit based on its attribute characteristics," can be achieved by executing... Figure 4 Steps 1021 to 1023 are implemented, and the details are explained below.

[0046] In step 1021, for each transaction data unit in the historical transaction data set, the merchant category identifier is matched based on a preset identifier mapping table, and the preset score corresponding to the matched merchant category identifier is determined as the basic score of the transaction data unit.

[0047] Here, the merchant category identifier is a set of standardized codes embedded in the transaction message, used to uniquely represent the industry attribute or business type of the resource recipient; the pre-built identifier mapping table is a pre-built key-value pair data structure that establishes a fixed mapping relationship between different industry codes and quantitative scores; the basic score is a necessary quantitative value initially determined based on industry attributes.

[0048] In actual implementation, the fixed-length field in the transaction data unit is read, the merchant category identifier is extracted, and a hash lookup algorithm is used to perform index matching in the identifier mapping table residing in memory. If the match is successful, the corresponding value is directly returned as the basic score; if the match fails (i.e., an undefined category identifier is encountered), the default scoring strategy is triggered.

[0049] In addition, if a perfect match fails, the default scoring strategy can adopt a fuzzy matching strategy, that is, extract the prefix field of the merchant category identifier (representing the major industry category) and perform a secondary search in the upper-level index of the mapping table, or call a pre-trained classifier model, take the unknown merchant category identifier as input feature, predict the major industry category to which it belongs and assign the corresponding average score, so as to ensure that effective scores can still be produced even if the identifier library is updated late.

[0050] As an example, suppose a transaction data unit The merchant category identifier carried in is (Corresponding to catering services), pre-set identifier mapping table The record contains {"5411":90, "5812":40, "4814":80}. Here, 90 points represents high necessity (e.g., supermarkets), and 40 points represents low necessity (e.g., dining out). The search matches... get The corresponding value of 40 represents the base score of the transaction data unit. It was determined to be 40 points.

[0051] In addition, a decision tree-based classification model can be used to replace the static mapping table. Merchant category identifiers and their associated geographic locations, transaction time periods, and other metadata are input into a pre-trained gradient boosting decision tree. The model outputs the predicted weights of the leaf nodes as the base score, thereby dynamically handling newly emerging or ambiguous merchant categories.

[0052] The above method matches merchant category identifiers with a pre-set identifier mapping table to determine the basic score. The hash lookup algorithm improves matching efficiency and ensures that the basic score is obtained efficiently and in a standardized manner. For undefined merchant category identifiers, the fuzzy matching strategy and pre-trained classifier model can effectively handle matching failure scenarios. Gradient boosting decision trees can dynamically adapt to newly emerging or fuzzy merchant categories, ensuring that a stable and effective basic score can still be output even when the identifier library is updated late. This significantly improves the reliability and applicability of the transaction data unit necessity quantification process.

[0053] In step 1022, target keywords are extracted from the transaction summary information, and the semantic score of the transaction data unit is determined based on the matching degree between the target keywords and each level category.

[0054] Here, transaction summary information is textual description data accompanying the transaction; the hierarchical keyword library is a knowledge base containing multiple hierarchical categories, each of which maintains a set of reference keywords composed of semantically representative words; and the matching degree is a normalized value that represents the relevance of the extracted content to a specific category.

[0055] In practice, the transaction summary information is parsed using a word segmentation algorithm, and after filtering out stop words, nouns or verbs with substantial meaning are selected as target keywords.

[0056] In some embodiments, Figure 4 The semantic score in step 1022 can be determined by the following method: obtaining a preset hierarchical keyword library, which includes a set of reference keywords corresponding to each hierarchical category; for each hierarchical category, counting the number of hits of the target keyword in the reference keyword set; determining the matching degree between the target keyword and the hierarchical category by the ratio of the number of hits to the total number of reference keywords in the reference keyword set; selecting the target hierarchical category with the highest matching degree from multiple hierarchical categories, and determining the preset score corresponding to the target hierarchical category as the semantic score of the transaction data unit.

[0057] In practical implementation, a pre-defined hierarchical keyword library is obtained. For each hierarchical category in the library, the reference keyword set for each category is traversed, and the number of times the target keyword appears in the reference keyword set is counted, recorded as the hit count. Next, a division operation is performed, dividing the hit count by the total number of keywords in the reference keyword set. The result is the matching degree between the target keyword and the hierarchical category. Finally, the matching degree values ​​calculated for all hierarchical categories are compared, and the target hierarchical category with the highest matching degree value is selected. The predefined score of the target hierarchical category in the configuration table is then determined as the semantic score of the transaction data unit.

[0058] In addition, to address the issue of missed matches caused by synonyms, a semantic expansion mechanism can be introduced before counting the number of hits. This can be done by using a thesaurus or a pre-trained word vector model to find synonyms for the target keyword and include these synonyms in the comparison with the reference keyword set. Alternatively, the cosine similarity between the target keyword vector and the reference keyword vector can be calculated, and a hit is considered when the similarity exceeds a threshold. A thesaurus is a database of synonyms containing a large number of synonym pairs.

[0059] For calculating the matching degree, a semantic vector space model can also be used. This involves encoding the transaction summary information into high-dimensional sentence vectors and encoding the standard descriptions of each category level into reference vectors. The cosine similarity between the sentence vectors and each reference vector is calculated, and the category score corresponding to the highest cosine similarity is used as the semantic score. This addresses the problem of inaccurate matching of synonyms (such as "pharmacy" and "drugstore").

[0060] As an example, suppose the transaction summary is "purchase of emergency medicines". The extracted target keywords are "emergency" and "medicines". Scanning the hierarchical keyword library, it is found that both words exist in the reference keyword set corresponding to the medical security category. The reference keyword set has ten reference words. At this time, the hit count is 2, and the corresponding matching degree is 0.2. If this matching degree is the highest among all categories, the target hierarchical category is determined to be the medical security category, and it is assigned a corresponding score of 95 as the semantic score.

[0061] By employing the above methods, a hierarchical keyword library is pre-set, and the matching degree is determined based on the ratio of the number of hits of the target keyword in the reference keyword set of each level category to the total number. The target level category with the highest matching degree is selected to assign a semantic score, which can accurately quantify the semantic attributes of transaction data units and improve the objectivity and stability of semantic scoring. By introducing a semantic expansion mechanism and a semantic vector space model, and by using synonym expansion and cosine similarity calculation, the problem of missed matching caused by synonyms can be effectively solved, enhancing the comprehensiveness and accuracy of keyword matching and avoiding semantic recognition bias caused by differences in expression. At the same time, by combining specific transaction summary examples, a feasible quantitative scoring can be achieved, ensuring that the semantic score calculation process is clear and the results are reliable, thereby improving the accuracy and applicability of semantic classification and scoring of transaction data as a whole.

[0062] In step 1023, the basic score and semantic score are weighted and summed according to the preset first weight ratio to obtain the necessity score of the transaction data unit.

[0063] Here, the first weighting ratio is a set of pre-set or dynamically generated coefficients used to balance the importance of structured and unstructured features in the final evaluation system; the necessity score is a comprehensive indicator used to determine the priority of resource return.

[0064] In actual implementation, the basic score output in step 1021 and the semantic score output in step 1022 are obtained. The basic score and the semantic score are multiplied by the corresponding weight coefficients in the first weight allocation, and then the two product results are added together to obtain the necessity score of the transaction data unit.

[0065] Furthermore, the allocation of the first weight can be designed as an adaptive adjustment mechanism, that is, calculating the confidence level (such as the degree of matching) during the semantic score generation process. If the degree of matching is extremely high, it indicates that the semantic information is very clear, and the weight of the semantic score can be automatically and temporarily increased. Conversely, if the semantic matching is too low, its weight will be automatically reduced, and more reliance will be placed on the basic score, thereby maintaining the robustness of the score under different data quality scenarios.

[0066] As an example, if the base score is 80 points and the weight is set to 0.6, and the semantic score is 95 points and the weight is set to 0.4, then the final necessity score is calculated as 80*0.6+95*0.4=86 points.

[0067] By using the above method, the basic score and semantic score are weighted and summed according to the preset first weight ratio to obtain the necessity score of the transaction data unit. This reasonably balances the roles of structured and unstructured features in the evaluation system. At the same time, an adaptive adjustment mechanism for the first weight is introduced to dynamically adjust the weight ratio based on the confidence level of the semantic score. This can maintain the robustness of the scoring under different data quality scenarios, making the necessity score more consistent with the actual data situation and improving the scientificity and reliability of resource return priority decision-making.

[0068] In some embodiments, hierarchical categories include at least survival categories, see further. Figure 3 , Figure 3 The step 102, “mapping multiple transaction data units to multiple hierarchical categories based on necessity scores,” can be achieved by performing the following method: obtaining a preset set of category division thresholds, which includes at least a first boundary threshold for defining survival categories; for each transaction data unit in the historical transaction data set, if the necessity score of the transaction data unit is greater than or equal to the first boundary threshold, mapping the transaction data unit to a survival category.

[0069] Here, the hierarchical categories are logical sets divided according to the urgency and necessity of resource outflows, and at least include survival categories that characterize the need to maintain the basic survival or core business operation of the target object. The category division threshold set is a set of pre-defined numerical boundaries used to divide the range of necessity scores into different intervals. Among them, the first dividing threshold is a key numerical indicator used to define the lower limit of the survival category. Transactions with scores reaching or exceeding the first dividing threshold are considered to have irreplaceable rigid needs.

[0070] In practice, the system reads the set of category classification thresholds, parses out the first dividing threshold, and then iterates through the historical transaction data set to extract the necessity score of each transaction data unit. For each transaction data unit, a numerical comparison logic is executed to determine whether its necessity score is greater than or equal to the first dividing threshold. If the comparison result is true, meaning the necessity score falls within the high-priority range, the corresponding transaction data unit is marked as belonging to the survival category, and a category identifier is written into the corresponding metadata field, completing the mapping from numerical value to category. If the comparison result is false, the system continues to compare against other thresholds in the set and classifies it into other levels of categories such as the development category or the enjoyment category.

[0071] In addition, a threshold adaptive adjustment mechanism based on dynamic programming or clustering algorithms can be introduced. This involves periodically statistically analyzing the distribution histogram of necessity scores in the historical transaction data set, finding the natural breakpoint of score density as the new first boundary threshold, or dynamically raising or lowering the threshold according to the current tension of the source resource pool. Specifically, when the available resources in the source resource pool are detected to be lower than the warning line, the first boundary threshold is automatically raised to narrow the coverage of the survival category, ensuring that only the most critical resource expenditures are classified as survival, thereby enabling strict management during periods of resource scarcity.

[0072] As an example, assuming the preset first threshold is 80 points, for a medical service transaction with a necessity score of 85 points, the comparison shows that 85 points is greater than the first threshold, so the transaction data unit is mapped to the survival category; while for a catering transaction with a score of 60 points, it is determined that it does not belong to the survival category.

[0073] By using the above method, and by pre-setting a set of category division thresholds and a first dividing threshold, transaction data units that meet the necessity score are mapped to the survival category. This can accurately distinguish between rigid needs and non-rigid expenditures, and clarify the priority of resource expenditures. The introduction of a threshold adaptive adjustment mechanism dynamically adjusts the dividing threshold based on the score distribution and resource pool status. This can shrink the coverage of the survival category when resources are scarce, improve the rationality and adaptability of hierarchical category division, ensure that resource allocation is more in line with actual operational needs, and enhance the scientific nature and stability of resource management.

[0074] In some embodiments, the set of category division thresholds further includes a second boundary threshold and a third boundary threshold, wherein the first boundary threshold is greater than the second boundary threshold, and the second boundary threshold is greater than the third boundary threshold; see also Figure 5 , Figure 5 This is a schematic diagram of the third process of the resource return processing method provided in the embodiments of this application; Figure 3 Step 102, "mapping multiple transaction data units to multiple hierarchical categories based on necessity scores," can also be performed by executing... Figure 5Steps 201 to 203 are implemented, and the details are explained below.

[0075] In step 201, for each transaction data unit in the historical transaction data set, if the necessity score of the transaction data unit is less than the first threshold and greater than or equal to the second threshold, the transaction data unit is mapped to the security category.

[0076] Here, the security category refers to resource expenditures that, while not essential for survival, play a vital role in mitigating future risks or maintaining the stability of life and business, such as insurance services or preventative maintenance; the second threshold is the numerical boundary that distinguishes security needs from development and improvement needs, and its value is set below the first threshold.

[0077] In practice, a dual-condition judgment instruction is adopted. First, the necessity score of the transaction data unit is read and compared with the first threshold. If the necessity score is less than the first threshold, the necessity score is then compared with the second threshold. When both conditions are met, the security category identifier is written into the metadata tag of the transaction data unit, mapping the transaction data unit to the security category.

[0078] In addition, a stability detection mechanism based on a sliding window can be introduced. When determining whether a transaction should be classified into a safe category, the stability of the transaction frequency over a past time window (such as six months) is calculated in addition to the single score. If the transaction shows high periodicity and fixed amount characteristics, additional stability points can be given, so that the score that was originally on the edge of the critical value can be stably placed above the second threshold, thereby avoiding the incorrect classification due to fluctuations in a single score.

[0079] As an example, suppose the first threshold is 80 points, the second threshold is 60 points, and the transaction score of the transaction data unit is 70 points. If the score of 70 points is within the range of 60 to 80, it will be classified as a safe category.

[0080] In step 202, if the necessity score of a transaction data unit is less than the second threshold and greater than or equal to the third threshold, the transaction data unit is mapped to the developmental category.

[0081] Here, the development category refers to resource inputs used to enhance personal skills, expand business scale, or increase future potential revenue, such as education and training or marketing; the third threshold is a numerical boundary used to define expenditures with value-added potential and purely consumable expenditures, and its value is set below the second threshold.

[0082] In practice, for transaction data units that fail to enter the survival or security categories, the necessity score is further compared with the third threshold. If the necessity score falls within the semi-closed interval formed by the second and third thresholds, the development category is selected as the mapping target, and the transaction data unit is mapped to the development category.

[0083] As an example, assuming the third threshold is 40 points, the transaction score of a transaction data unit is 50 points, and the transaction score is less than 60 points but greater than 40 points, then it is identified and mapped to the developmental category.

[0084] In step 203, if the necessity score of a transaction data unit is less than the third threshold, the transaction data unit is mapped to the enjoyment category.

[0085] Here, the "enjoyment-type" category represents non-essential, non-value-adding resource consumption behaviors with high substitutability, typically corresponding to entertainment scenarios.

[0086] In practice, transaction data units that fail to meet all the aforementioned threshold conditions, i.e., the remaining items with scores below the third threshold, will be automatically mapped to the enjoyment category.

[0087] In addition, anomaly detection and behavior correction prompts can be integrated in step 203. Clustering algorithms can be used to subgroup transaction data classified into the enjoyment category to identify high-frequency and inefficient long-tail expenditures (such as daily repeated purchases of expensive beverages). The total amount of resources that can be released if this part of the expenditure is reduced can be calculated, and a specific resource optimization prediction model can be generated and stored in the database to provide data support for subsequent resource return recommendations.

[0088] As an example, a transaction data unit with a transaction score of 30 is explicitly mapped to the enjoyment category because its transaction score is below the third threshold of 40.

[0089] By using the above method, and through the first, second, and third boundary thresholds, transaction data units are sequentially mapped to the security, development, and enjoyment categories. This enables multi-level and precise classification of transaction data, clearly defining the attribute positioning of different expenditures. Employing dual-condition judgment instructions improves the accuracy of category attribution, and introducing a sliding window stability detection mechanism avoids classification errors caused by single score fluctuations. By using clustering algorithms to detect anomalies in the enjoyment category, inefficient expenditures can be identified and a resource optimization prediction model can be generated. This provides reliable data support for resource return recommendations and improves the standardization and rationality of resource management.

[0090] See also Figure 3In step 103, based on the transaction data units belonging to the survival category, the proportion of survival transactions is determined, and the resource pressure level of the target resource account is determined based on the proportion of survival transactions.

[0091] In some embodiments, see Figure 6 , Figure 6 This is a schematic diagram of the fourth process of the resource return processing method provided in the embodiments of this application; Figure 3 Step 103, "Determine the proportion of survival-type transactions based on transaction data units belonging to the survival-type category," can be achieved by executing... Figure 6 Steps 1031 to 1033 are implemented, and the details are explained below.

[0092] In step 1031, the resource quantities of transaction data units belonging to the survival category are summed to obtain the first total resource quantity corresponding to the survival category.

[0093] Here, resource quantity is a numerical metric attribute recorded in the transaction data unit, and the first total resource quantity is a statistical indicator representing the absolute scale of rigid demand.

[0094] In practice, an accumulator variable is initialized to zero. Then, the historical transaction data set is traversed, and the category label of each transaction data unit is checked. If the category label shows a survival category, the resource quantity field value of the transaction data unit is read and added to the accumulator. After the traversal is completed, the final value of the accumulator is the first total resource quantity.

[0095] As an example, suppose the historical transaction data set contains three transaction data units belonging to the survival category, with resource quantities of 100, 200 and 50 respectively, and the first total resource quantity is calculated to be 350.

[0096] In step 1032, the resource amount of each transaction data unit in the historical transaction data set is summed to obtain the second total resource amount corresponding to the historical transaction data set.

[0097] In practice, a high-efficiency scanning mechanism of columnar storage database is adopted, which directly performs a full summation operation on the column data of the storage resource quantity, or maintains a global total resource quantity counter through triggers when transaction data is written to the database, and directly reads the real-time value of the counter as the second total resource quantity, thereby avoiding the overhead of full table scan.

[0098] As an example, suppose the historical transaction data set contains other types of transactions with a total resource amount of 150 in addition to the aforementioned survival transactions. Then, by adding up the resource amounts of all transactions, the second total resource amount is 500.

[0099] In step 1033, the ratio of the first total resource quantity to the second total resource quantity is determined as the survival transaction proportion.

[0100] In practice, a normalization process is performed, and the ratio of the first total resource quantity to the second total resource quantity is determined as the survival-oriented transaction proportion. Here, the survival-oriented transaction proportion is a dimensionless percentage value used to measure the structural weight of rigid resource demand in the overall resource allocation.

[0101] As an example, based on the aforementioned calculation results, 350 / 500=0.7, that is, 70% of the calculation results, is taken as the proportion of survival-type transactions.

[0102] By summing the resource amounts of survival-type transaction data units as described above, we obtain the first total resource amount. By summing the resource amounts of all transaction data units in the historical transaction data set, we obtain the second total resource amount. The ratio of the two is then used to determine the proportion of survival-type transactions. This method can objectively quantify the structural weight of rigid demand in the overall resource allocation. The use of an efficient summation mechanism can reduce computational overhead and improve statistical efficiency. The calculation process is standardized and the results are intuitive and accurate, providing a reliable basis for subsequently determining the resource pressure level of target resource accounts.

[0103] In some embodiments, see Figure 7 , Figure 7 This is a schematic diagram of the fifth process of the resource return processing method provided in the embodiments of this application; Figure 3 Step 103, "Determining the resource pressure level of the target resource account based on the proportion of survival-oriented transactions," can be achieved through execution... Figure 7 Steps 1034 to 1038 are implemented, and the details are explained below.

[0104] In step 1034, the periodic resource inflow of the target resource account and the number of historical overdue payments within a preset statistical period are obtained.

[0105] Here, periodic resource inflow refers to the stable and recurring resource replenishment value obtained by the target resource account per unit of time; historical overdue number refers to the expected number of records in which the target resource account fails to fulfill its resource return obligation within the preset statistical period; the preset statistical period is a time window used to evaluate historical behavior, such as the past twelve months.

[0106] In actual implementation, the system calls the resource inflow records of the target resource account's management system through the application programming interface, extracts the inflow-type resource data, and calculates the mean or median as the periodic resource inflow amount; at the same time, it scans the overdue logs, counts the number of events marked as overdue, and determines the historical overdue number within the preset statistical period.

[0107] As an example, if the target account has a monthly resource inflow of 10,000 and there have been two instances of overdue resource returns in the past twelve months, then the values ​​of 10,000 and the two instances are used as the basis for subsequent calculations.

[0108] In step 1035, the ratio of the second total resource quantity to the periodic resource inflow is determined as the resource load ratio of the target resource account.

[0109] Here, the second total resource quantity, the historical total resource outflow statistically calculated in the aforementioned steps, and the resource load ratio directly reflect whether the inflow of resources covers the outflow and the degree of overspending.

[0110] In actual implementation, the total outflow obtained in step 1032 is divided by the inflow obtained in step 1034 to obtain a floating-point number reflecting the resource burden. If the floating-point ratio is greater than 1, it means that the company is in a deficit state; if the floating-point ratio is less than 1, it means that the company is in a surplus state.

[0111] As an example, if the second total resource quantity is 8000 and the periodic resource inflow is 10000, then the resource load ratio is 8000 / 10000=0.8, indicating that the current resource consumption is within a controllable range.

[0112] In step 1036, the ratio of the number of historical overdue payments to the preset statistical period is determined as the normalized value of the number of historical overdue payments corresponding to the target resource account.

[0113] In practice, the historical overdue number counts calculated in step 1034 are read and divided by the number of time units (such as months) included in the preset statistical period. The quotient is the normalized value of the historical overdue number counts. The normalized value of the historical overdue number counts is an indicator that characterizes the frequency density of overdue behavior, eliminating the influence of the length of the preset statistical period on the absolute value of the overdue number counts. The preset statistical period is a pre-set time span used to observe and statistically analyze historical behavior data. It is usually determined according to business retrospective needs and is used to define the time window boundaries for risk assessment, such as the most recent twelve months or the most recent twenty-four weeks.

[0114] In addition, a time decay function can be introduced when calculating the normalized value to assign different weights to overdue behaviors that occur at different time points. Overdue behaviors that are closer to the current time point are assigned a larger weight coefficient, while those that are farther away are assigned a smaller weight coefficient. By weighted summation and then dividing by the period length, the normalized value can more sensitively reflect the recent risk trend of the target resource account.

[0115] As an example, if two overdue payments occur within a twelve-month statistical period, the normalized value of the historical overdue number is approximately 0.167.

[0116] In step 1037, the resource stress index of the target resource account is obtained by weighting the proportion of survival-type transactions, the resource load ratio, and the normalized value of historical overdue times according to the preset second weighting ratio.

[0117] Here, the resource stress index is a composite evaluation indicator that comprehensively reflects the degree of resource burden and the risk of overdue payments; the second weighting ratio is a set of pre-set coefficient vectors, which correspond to the importance of the aforementioned three indicators in the comprehensive evaluation.

[0118] In actual implementation, the survival transaction ratio, resource load ratio, and normalized value of historical overdue number obtained from the previous steps are read respectively. These three values ​​are multiplied by the corresponding weight coefficient in the second weight allocation, and then the three product results are added together. The sum is the resource tension index.

[0119] As an example, if the proportion of survival-type transactions is 0.6, the resource load ratio is 0.8, the normalized value of historical overdue number is 0.1, and the weights are 0.4, 0.4, and 0.2 respectively, then the calculated resource stress index is 0.58.

[0120] In step 1038, the resource stress level of the target resource account is determined based on the resource stress index.

[0121] In some embodiments, Figure 7 Step 1038 can be achieved by the following methods: obtaining a preset first stress threshold and a preset second stress threshold; if the resource stress index is greater than the first stress threshold, determining the resource pressure level of the target resource account as high stress level; if the resource stress index is less than or equal to the first stress threshold and the resource stress index is greater than or equal to the second stress threshold, determining the resource pressure level of the target resource account as medium stress level; if the resource stress index is less than the second stress threshold, determining the resource pressure level of the target resource account as low stress level.

[0122] Here, the resource stress level is a classification label after discretizing the resource tension index, used to indicate the severity of the resource management strategies to be adopted subsequently, and is divided into high stress level, medium stress level and low stress level.

[0123] In actual implementation, a preset first tension threshold and a preset second tension threshold are applied, with the first tension threshold being greater than the second tension threshold. Then, the calculated resource tension index is compared with these two thresholds. If the resource tension index is greater than the first tension threshold, it is determined to be a high-pressure level; if the resource tension index is less than or equal to the first tension threshold but greater than or equal to the second tension threshold, it is determined to be a medium-pressure level; if the resource tension index is less than the second tension threshold, it is determined to be a low-pressure level.

[0124] In addition, a hysteresis comparator mechanism is introduced into the threshold determination logic. A buffer zone is set near the first tension threshold and the second tension threshold respectively. When the resource tension index falls into the buffer zone, the resource pressure level of the previous moment remains unchanged until the index completely breaks through the boundary of the buffer zone before the level is switched. This avoids frequent changes in pressure level due to small fluctuations of the index near the threshold and ensures stability.

[0125] By using the above method, and by setting a first stress threshold and a second stress threshold, the resource stress index is compared with the threshold, which can accurately classify the high, medium and low resource pressure levels of the target resource account, providing a clear classification basis for subsequent resource management strategies. The introduction of a hysteresis comparator mechanism and the setting of a buffer zone can effectively avoid the frequent jumps in stress level caused by fluctuations in the resource stress index around the threshold, improve the stability and reliability of the level determination, and ensure the continuous and stable resource management process.

[0126] See also Figure 3 In step 104, the comprehensive score of each transaction data unit is determined based on the necessity score and transaction attributes of each transaction data unit.

[0127] In some embodiments, see Figure 8 , Figure 8 This is a schematic diagram of the sixth process of the resource return processing method provided in the embodiments of this application; Figure 3 Step 104 in the process can be executed Figure 8 Steps 1041 to 1045 are implemented, and the details are explained below.

[0128] In step 1041, the necessity score of each transaction data unit is normalized to obtain the necessity weight of the transaction data unit.

[0129] Here, the necessity weight is a normalized value that reflects the importance of the transaction data unit to maintaining the survival or basic operation of the target object, and the value range is usually between 0 and 1; the necessity score is the original score assigned in the previous steps based on the hierarchical category to which the transaction belongs (such as survival type, development type, etc.).

[0130] In practice, all transaction data units to be processed are traversed to obtain their respective necessity scores, and the maximum and minimum values ​​are found. The deviation standardization algorithm is used to subtract the minimum value from the necessity score of each transaction data unit, and then divide by the difference between the maximum and minimum values. The result is the necessity weight of the transaction data unit.

[0131] As an example, if the transaction necessity score of a certain transaction data unit is 90 points, and the highest transaction necessity score among all transaction data units is 100 points and the lowest score is 10 points, after linear normalization, the necessity weight of the transaction data unit 90-10 / 100-90 is approximately 0.89.

[0132] By using the above method, the necessity score of the transaction data unit is normalized through the deviation standardization algorithm, and a necessity weight with a value between 0 and 1 can be obtained. This objectively reflects the importance of each transaction data unit, and the numerical values ​​are uniform and easy to calculate in the future. The processing is standardized and stable, which can effectively eliminate the difference in the magnitude of the score and improve the accuracy and rationality of subsequent resource allocation and decision-making.

[0133] In step 1042, the urgency weight of the transaction data unit is determined based on the transaction generation time and settlement reference time of the transaction data unit.

[0134] In some embodiments, Figure 8 Step 1042 can be achieved by: determining the time difference between the settlement reference time and the transaction generation time of the transaction data unit; and determining the urgency weight of the transaction data unit as the ratio of the time difference to the settlement cycle duration corresponding to the transaction data unit.

[0135] Here, the transaction generation time refers to the timestamp of the transaction instruction generation or billing date; the settlement reference time refers to the deadline or agreed return date for the resources to be returned; and the urgency weight is a timeliness indicator that quantifies the priority of transaction processing.

[0136] In practice, the time difference between the settlement reference time of the transaction data unit and the transaction generation time is calculated, the preset settlement cycle length of the transaction is obtained (for example, the monthly cycle length is thirty days), the time difference is divided by the settlement cycle length, and the resulting ratio is used as the urgency weight of the transaction data unit.

[0137] Furthermore, considering the non-linear impact of time elapsed on urgency, a dynamic remaining time factor is introduced when calculating urgency weights. The current system time is used instead of the transaction generation time in the calculation, that is, the difference between the settlement reference time and the current time is calculated. As the current time approaches the settlement reference time, this difference decreases. This difference is processed through an inverse proportional function or an exponential decay function, so that the urgency weight of transactions with shorter remaining time increases exponentially, thereby ensuring that near-expiration transactions can be prioritized.

[0138] By using the above method, the urgency weight is determined by the time difference between the settlement reference time and the transaction generation time, as well as the settlement cycle length, which can quantify the timeliness priority of transaction processing. Introducing a dynamic remaining time factor and an exponential decay function can significantly increase the urgency weight of near-expiration transactions, ensuring that transactions nearing their return deadline are processed first, and improving the rationality and timeliness of resource return scheduling.

[0139] In step 1043, the ratio of the resource quantity of the transaction data unit to the total resource quantity of the historical transaction data set is determined as the resource quantity percentage of the transaction data unit.

[0140] Here, the resource volume ratio is a value reflecting the contribution of a single transaction data unit to the overall resource return; the total resource volume of the historical transaction data set refers to the total resources of all pending or processed transactions within the statistical period.

[0141] In actual implementation, the resource quantity value of the current transaction data unit is read, and the database aggregation function is called to calculate the total resource quantity of all pending transaction data in the current batch. The division operation is performed to obtain the ratio, and the ratio is determined as the resource quantity proportion of the transaction data unit.

[0142] In step 1044, the preference coefficient corresponding to the transaction data unit is determined based on the historical resource return behavior of the target resource account.

[0143] In some embodiments, Figure 8 Step 1044 can be implemented by the following method: detecting whether there are valid historical resource return records in the target resource account and obtaining the detection result; if the detection result indicates that the target resource account has historical resource return records, calculating the frequency at which transactions belonging to the hierarchical category of the transaction data unit are prioritized, and using this as the preference coefficient corresponding to the transaction data unit; if the detection result indicates that the target resource account does not have historical resource return records, matching the corresponding default preference value from the preset cold start preference mapping table according to the hierarchical category of the transaction data unit as the preference coefficient corresponding to the transaction data unit.

[0144] Here, the preference coefficient is a personalized weight that reflects the target resource account's tendency to process specific types of transactions; the historical resource return record is a sequential log of the target resource account's returns of different types of transactions in the past when resources were sufficient or insufficient.

[0145] In actual implementation, the database is first queried to check whether there are valid historical resource return records for the target resource account. If the detection result indicates that there are records, the frequency of transactions belonging to the hierarchical category of the current transaction data unit (e.g., catering, entertainment) in the historical data is counted, and the frequency of pre-processing is used as the preference coefficient. If the detection result indicates that there are no records, i.e., in the case of a cold start scenario, the hierarchical category of the transaction data unit is identified, and the preset cold start preference mapping table is accessed. The cold start preference mapping table stores default weights based on common public behavior. The corresponding default preference value is matched from the cold start preference mapping table and used as the preference coefficient.

[0146] In addition, a time decay mechanism can be introduced when calculating historical preferences, giving higher statistical weight to historical return records within a preset time period and lower weight to records older than the preset time period, in order to adapt to the dynamic changes in consumption habits and return preferences. At the same time, for users without records, clustering algorithms can be used to classify them into groups with similar behavioral characteristics, and the average preference value of the group can be used to replace the static mapping table values.

[0147] By using the above method, the preference coefficient is determined based on the historical resource return behavior of the target resource account, which can reflect the personalized processing tendency. When there is a historical record, the priority processing frequency is statistically analyzed. When there is no record, the default value is obtained through the cold start preference mapping table. By introducing a time decay mechanism and clustering algorithm, it can adapt to changes in user preferences and improve the accuracy and applicability of the preference coefficient.

[0148] In step 1045, the necessity weight, urgency weight, resource quantity ratio and preference coefficient are weighted and summed according to the preset third weight ratio to obtain the comprehensive score of the transaction data unit.

[0149] Here, the overall score is a unique scalar value used to determine the order of transaction data units in the resource return queue; the third weighting ratio is a set of coefficient vectors that sum to 1, which respectively define the relative importance of the above four dimensions in the overall evaluation.

[0150] In actual implementation, a third weighting ratio is loaded (for example, the necessity weight ratio is 40%, the urgency weight ratio is 30%, the resource quantity ratio is 20%, and the preference coefficient ratio is 10%). The necessity weight, urgency weight, resource quantity ratio, and preference coefficient obtained in steps 1041 to 1044 are multiplied by the corresponding ratio coefficients, and the four product results are summed to output the comprehensive score.

[0151] By normalizing the necessity score to obtain the necessity weight, combining the transaction time information to determine the urgency weight, calculating the proportion of transaction resources, determining the preference coefficient based on historical resource return behavior, and then weighting and summing according to the preset third weight ratio to obtain the comprehensive score, the transaction data unit can be comprehensively evaluated from multiple dimensions. The numerical values ​​are standardized and the logic is rigorous, which can accurately determine the order of resource return queues and improve the scientificity, rationality and personalization of resource return scheduling.

[0152] See also Figure 3 In step 105, when the available resources are less than the total resources of the historical transaction data set, a combined resource return strategy containing segmented return instructions is generated based on the resource pressure level and the comprehensive score of each transaction data unit.

[0153] In some embodiments, see Figure 9 , Figure 9 This is a schematic diagram of the seventh process of the resource return processing method provided in the embodiments of this application; Figure 3 Step 105, "Generating a combined resource return strategy containing segmented return instructions based on resource pressure levels and the comprehensive score of each transaction data unit," can be executed... Figure 9 Steps 1051 to 1052 are implemented, and the details are explained below.

[0154] In step 1051, based on the comprehensive score of each transaction data unit, multiple transaction data units in the historical transaction data set are sorted to obtain a transaction data unit sequence.

[0155] Here, the overall score is a composite quantitative indicator calculated in the preceding steps, reflecting the necessity, urgency, resource consumption, and preferences of the transaction; the transaction data unit sequence is an ordered list of data arranged in order of priority from high to low or from low to high.

[0156] In practice, quicksort or mergesort algorithms are used, with the overall score as the sorting key, to sort all transaction objects in the historical transaction data set in descending order. If there are transaction data units with the same overall score, their necessity weight or urgency weight is further compared to determine the final order.

[0157] For example, if the overall score of transaction data unit A is 0.9, the overall score of transaction data unit B is 0.6, and the overall score of transaction data unit C is 0.8, then the generated sequence of transaction data units is A, C, B.

[0158] By using the above method, discrete evaluation results are transformed into a linear execution sequence, ensuring that the resource allocation process strictly follows the established evaluation logic and providing orderly data input for subsequent strategy generation.

[0159] In step 1052, a combined resource return strategy containing segmented return instructions is generated based on the resource pressure level and the transaction data unit sequence.

[0160] Here, the segmented repayment instruction includes at least one of the following: minimum repayment instruction, full repayment instruction, and installment repayment instruction.

[0161] In some embodiments, see Figure 10 , Figure 10 This is the eighth flowchart of the resource return processing method provided in the embodiments of this application; Figure 9 Step 1052 in the process can be executed Figure 10 Steps 301 to 304 are implemented, and the details are explained below.

[0162] In step 301, the total amount of resources and the minimum amount of resources to be returned corresponding to the historical transaction data set are determined.

[0163] Here, total resources refer to the sum of resources from all pending transactions in the current period, representing all resource repayment obligations faced by the target resource account; minimum resource repayment refers to the minimum resource threshold that must be repaid to maintain the normal credit status of the account according to business rules or contractual agreements, which is usually a certain percentage of total resources.

[0164] In actual implementation, the historical transaction data set is traversed, and the resource amount values ​​of all transaction data units are accumulated to obtain the total resource amount. Then, the minimum return ratio coefficient preset in the configuration parameters (e.g., 10%) is read, the total resource amount is multiplied by the minimum return ratio coefficient, and the minimum absolute value lower limit is referenced to calculate the minimum return resource amount.

[0165] For example, if the total resource quantity is 10,000 and the minimum return ratio is 10%, then the minimum resource quantity to be returned is 1,000. This method establishes two key boundary values ​​for resource return decisions, providing a quantitative benchmark for subsequently assessing the sufficiency of available resources and selecting the appropriate strategy.

[0166] In step 302, when the available resources are greater than or equal to the total resources, a first type of combined resource return strategy is generated based on the resource pressure level and the transaction data unit sequence.

[0167] In some embodiments, Figure 10 Step 302 can be achieved by generating a first type of combined resource return strategy that includes a minimum amount return instruction when the resource pressure level is high.

[0168] Here, when there is a target transaction data unit in the transaction data unit sequence with a resource quantity greater than the preset resource quantity and a necessity score greater than the preset score threshold, the first type of combined resource return strategy includes a single installment return instruction for the target transaction data unit.

[0169] In practical implementation, under high resource pressure, this means that although the current available resources are sufficient to cover the entire repayment amount, defensive planning is needed based on future overdue pressure. If there is a target transaction data unit in the transaction data unit sequence with a resource amount greater than the preset resource amount and a necessity score greater than the preset score threshold, a strategy including a minimum repayment instruction is generated to preserve future resource liquidity. Simultaneously, for target transaction data units with a necessity score greater than the preset score threshold, a single installment repayment instruction for that target transaction data unit will be attached to the first type of combined resource repayment strategy. Conversely, the first type of combined resource repayment strategy includes a minimum repayment instruction.

[0170] For example, under a high-pressure resource situation, the corresponding combined resource repayment strategy for the first type is a minimum repayment instruction. Further, it is determined whether there are any target transaction data units with a resource quantity greater than a preset resource quantity and a necessity score greater than a preset score threshold. For example, if the preset resource quantity is 8000, and a transaction data unit with a necessity score of 95 and a resource quantity of 10000 exists, then a single installment repayment instruction for that transaction data unit is added to the combined resource repayment strategy. If none of the above situations exist, only a recommended minimum repayment instruction is included.

[0171] When the resource pressure level is medium, a first-type combined resource repayment strategy containing installment repayment instructions is generated.

[0172] Here, when the resource proportion of transaction data units belonging to the enjoyment category in the transaction data unit sequence is greater than the first proportion threshold, the installment repayment instruction corresponds to the first installment period; when the resource proportion is between the second proportion threshold and the first proportion threshold, the installment repayment instruction corresponds to the second installment period.

[0173] In actual implementation, if the resource pressure level is medium pressure level, the resource ratio of transaction data units belonging to the enjoyment category is calculated. If the resource ratio exceeds the first ratio threshold, an instruction for the corresponding first installment period is generated. If the resource ratio is between the second ratio threshold and the first ratio threshold, an instruction for the corresponding second installment period is generated. The value of the first installment period is greater than the value of the second installment period, which is used to curb the consumption of the enjoyment category from occupying the available resources.

[0174] For example, if the first proportion threshold is set to 40% and the second proportion threshold is set to 30%, and the calculated resource proportion exceeds 40%, then an instruction corresponding to the first installment period is generated, such as recommending a six-installment repayment and providing transaction prompts for non-essential transactions. If the resource proportion is between 30% and 40%, then an instruction corresponding to the second installment period is generated, such as recommending a three-installment repayment and providing transaction prompts for non-essential transactions. Meanwhile, if the resource proportion is less than the second proportion threshold, then an instruction for the second installment period can be directly recommended, and there is no need to provide transaction prompts for non-essential transactions.

[0175] When the resource pressure level is low, generate a first type of combined resource return strategy that includes a full return instruction.

[0176] In actual implementation, if the resource pressure level is low, a strategy containing a full refund instruction will be generated directly to avoid additional resource overhead. It can also provide a good indication of the distribution of consumption levels and the amount of available resources, without the burden of resource refund.

[0177] By determining the total amount of resources and the minimum amount of resources to be returned, a quantitative benchmark is provided for resource return decisions. Then, by combining different resource pressure levels and transaction data unit sequences, a combined resource return strategy with corresponding segmented return instructions is generated. The minimum amount, installment, or full return instructions can be flexibly selected according to the pressure level. At the same time, the number of installments can be adjusted according to the proportion of enjoyment-type resources. This can not only protect the credit status of the account, but also reasonably control resource occupation and improve the adaptability and scientific nature of resource return arrangements.

[0178] In step 303, when the available resources are less than the total resources but greater than or equal to the minimum resource return amount, a second type of combined resource return strategy is generated based on the multidimensional evaluation model.

[0179] In some embodiments, the hierarchy further includes: a security category, a development category, and an enjoyment category; see also Figure 11 , Figure 11 This is a ninth flowchart illustrating the resource return processing method provided in this application embodiment; Figure 10 Step 303 in the process can be executed Figure 11Steps 3031 to 3035 are implemented, and the details are explained below.

[0180] In step 3031, if the historical overdue normalized value of the target resource account within a preset statistical period is greater than or equal to a preset risk threshold, a second type of combined resource repayment strategy containing a full repayment instruction is generated.

[0181] In practice, the second type of combined resource repayment strategy is a refined scheduling scheme for scenarios with limited available resources. The historical overdue normalized value is an indicator reflecting the frequency of overdue payments calculated in the previous steps; the preset risk threshold is the critical point of credit risk. The historical overdue normalized value of the target resource account is read. If the historical overdue normalized value is greater than or equal to the preset risk threshold, it indicates that the target resource account's recent credit record has deteriorated. To prevent further credit collapse, a second type of combined repayment strategy is generated that covers as much of the resources to be repaid as possible. This means that the user is advised to use backup resources for full repayment to repair their credit record.

[0182] For example, if the target resource account has a history of frequent overdue payments, and despite the current shortage of available resources, the risk is deemed too high, it is recommended to repay the full amount to avoid account freezing.

[0183] In step 3032, if the historical overdue normalized value is less than the preset risk threshold, the transaction rationality index of the target resource account is obtained by weighting and summing the proportion of survival-type transactions, the proportion of safe-type transactions, the proportion of development-type transactions, and the proportion of enjoyment-type transactions in the historical transaction data set according to the preset fourth weight ratio.

[0184] In practice, the transaction rationality index is a quantitative indicator for evaluating the health of a user's resource transaction structure. The fourth weighting ratio is a combination of coefficients, typically giving higher positive weights to survival and development types, and lower or negative weights to enjoyment types. Specifically, the proportion of the four types of transactions in the historical transaction data set is first calculated, and then multiplied by the corresponding weighting coefficients (e.g., survival type 0.41, security type 0.3, development type 0.2, enjoyment type 0.1). The sum is then used to obtain an index between 0 and 1. The higher the value, the more the resource transaction structure tends to be based on basic needs and value-added, and the higher the rationality of the phased transactions.

[0185] In addition, an objective weighting model based on entropy weighting can be used to dynamically adjust the weights of various categories according to the current overall distribution of resource transaction data in society. For example, during an economic downturn, the weight of survival-related transactions can be further increased, so that the calculated rationality index is more in line with the value judgment under the macro environment.

[0186] In step 3033, the resource difference between the available resource quantity and the minimum repaid resource quantity corresponding to the historical transaction data set is determined, and the ratio of the resource difference to the necessary resource expenditure of the target resource account in a preset period is determined as the liquidity buffer rate of the target resource account.

[0187] In practice, the liquidity buffer ratio of the target resource account is determined by comparing the available resource amount with the minimum repaid resource amount corresponding to the historical transaction data set. This difference is then used as the ratio of the resource difference to the necessary resource expenditure of the target resource account within a preset period. The liquidity buffer ratio reflects the ability of the remaining available resources to support future rigid transaction data after the user has repaid the minimum repaid resource amount. The necessary resource expenditure is the predicted amount of resources required for the user to survive in the next cycle. Specifically, the liquidity buffer ratio is calculated by subtracting the minimum repaid resource amount from the available resources, and simultaneously predicting the necessary resource expenditure within a preset period based on historical transaction data. The difference is then divided by the predicted necessary resource expenditure, and the resulting ratio is the liquidity buffer ratio. The preset period is the time frame corresponding to the next future cycle.

[0188] In step 3034, the phased return score of the target resource account is determined based on the transaction rationality index and the liquidity buffer ratio.

[0189] Here, the installment income score is positively correlated with the transaction rationality index and negatively correlated with the liquidity buffer ratio.

[0190] In practice, the installment return score is an indicator that measures the overall utility brought to users by the installment strategy. Specifically, a multivariate function can be constructed, setting the installment return score to be directly proportional to the transaction rationality index and inversely proportional to the liquidity buffer rate. For example, it can be achieved by the formula: Transaction Rationality Index × Rationality Coefficient + (1 - Liquidity Buffer Rate / Preset Buffer Base) × Buffer Rate Coefficient. Here, the rationality coefficient is used to measure the degree of importance attached to the rationality of the target resource account's past resource usage structure, the buffer rate coefficient is used to measure the degree of emphasis on alleviating the current liquidity pressure of the target resource account in the generation of the resource return strategy, and the preset buffer base is a preset percentage value used to measure whether the liquidity buffer rate is in a safe range. It is the minimum liquidity surplus ratio that a healthy target resource account should retain.

[0191] Furthermore, the calculation process can incorporate a nonlinear utility function, where the phased return score increases exponentially when the liquidity buffer ratio falls below a certain threshold. This couples the rationality of the transaction data structure and the degree of resource availability into a unified decision variable, solving the problem that a single dimension cannot comprehensively assess the necessity of phased allocation.

[0192] In step 3035, a second type of combined resource return strategy is generated based on the transaction rationality index, liquidity buffer ratio, and installment income score using a preset decision matrix.

[0193] In some embodiments, see Figure 12 , Figure 12 This is a schematic diagram of the tenth process of the resource return processing method provided in the embodiments of this application; Figure 11 Step 3035 in the process can be executed Figure 12 Steps 30351 to 30356 are implemented, and the details are explained below.

[0194] In step 30351, the total installment service fee is determined based on the total resource quantity corresponding to the historical transaction data set, the preset unit period service fee rate, and the preset number of installment periods, and the ratio of the total installment service fee to the total resource quantity is determined as the installment fee ratio.

[0195] Here, if the installment benefit score is greater than the installment fee ratio, it means that installment repayment has a positive benefit.

[0196] In practice, determining the total installment service fee based on the total resource volume corresponding to the historical transaction data set, the preset unit period service fee rate, and the preset number of installment periods can be achieved through a multiplicative-accumulator algorithm. Here, the total resource volume refers to the total principal amount that the target resource account should repay within the current repayment period; the unit period service fee rate is a preset time value coefficient used to cover the resource occupancy cost; and the preset number of installment periods is the number of standard time slices. After obtaining the total installment service fee through multiplication, the ratio of the total installment service fee to the total resource volume is calculated. This ratio is determined as the installment fee ratio, which characterizes the unit cost of the installment activity.

[0197] Based on this, the installment revenue score calculated in the previous steps is compared with the installment fee ratio. If the comparison result indicates that the installment revenue score is greater than the installment fee ratio, it is determined that the installment repayment at this time has positive revenue. That is, the value of the improvement in ecological health (e.g., maintaining the activity of the target resource account) brought about by providing installment services to the target resource account is higher than the marginal cost of resource occupation.

[0198] In step 30352, when the transaction rationality index indicates high rationality and the liquidity buffer ratio indicates high pressure, a second type of combined resource repayment strategy containing long-term installment repayment instructions is generated.

[0199] In practice, when the transaction rationality index indicates high rationality and the liquidity buffer ratio indicates high pressure, it means that although the target resource account has a healthy historical transaction structure (mostly used for production or essential consumption), its current available liquidity resources are extremely scarce, facing the risk of overdue payments. In this case, a second type of combined resource repayment strategy, including long-term installment repayment instructions, is generated. Here, a high rationality transaction rationality index means the index value is higher than a preset first threshold (e.g., 0.8); a high pressure liquidity buffer ratio usually means the ratio is lower than a preset warning threshold (e.g., 0.1). Long-term installment repayment instructions refer to setting the preset installment period to the maximum allowed value (e.g., 24 or 36 periods), thereby maximizing the repayment timeline and reducing the resource amount required for repayment in a single cycle.

[0200] In step 30353, when the transaction rationality index indicates high rationality, the liquidity buffer ratio indicates moderate pressure, and installment repayment has positive returns, a second type of combined resource repayment strategy containing medium- and long-term installment repayment instructions is generated.

[0201] In practice, a moderate stress state refers to a liquidity buffer ratio between the first and second thresholds (e.g., between 0.1 and 0.3). When the transaction rationality index indicates high rationality, the liquidity buffer ratio indicates moderate stress, and installment repayment has positive returns, although installment repayment has positive returns, considering that the target resource account has a certain resource repayment capacity, the number of installment periods is set to a moderate length (e.g., 12 periods). This generates a second type of combined resource repayment strategy that includes medium- and long-term installment repayment instructions. This alleviates the resource repayment pressure on the target resource account while avoiding excessive long-term resource occupation, achieving a balance between risk mitigation and resource turnover efficiency.

[0202] In step 30354, when the transaction rationality index indicates high rationality, the liquidity buffer ratio indicates low pressure, and the phasing has positive returns, a second type of combined resource repayment strategy containing excess repayment instructions is generated.

[0203] Here, the overpayment instruction indicates a repayment amount higher than the minimum repayment amount.

[0204] In practice, when the transaction rationality index indicates high rationality, the liquidity buffer ratio indicates low pressure, and the installment has positive returns, even if the installment return score is positive, it is usually recommended that the target resource account return the full amount or a large proportion of the resources because the available resources in the target resource account are sufficient. This generates a second type of combined resource return strategy that includes an over-return instruction. The over-return instruction guides the target resource account to use the liquidity of idle available resources to reduce future additional resource expenditures.

[0205] For example, for a target resource account with sufficient available resources, the default installment plan will no longer be recommended. Instead, an instruction will be sent suggesting that 80% of the resources to be repaid be returned.

[0206] In step 30355, when the transaction rationality index indicates moderate rationality and the liquidity buffer ratio indicates high pressure, a second type of combined resource repayment strategy is generated, which includes a minimum amount repayment instruction and a funding reminder.

[0207] In practice, when the transaction rationality index indicates moderate rationality and the liquidity buffer ratio indicates high pressure, the target resource account is determined to be on the edge of risk and its resource utilization efficiency is average. The second type of combined resource repayment strategy generated at this time includes a minimum repayment instruction and resource sourcing prompts, such as part-time job referrals or the conversion of idle, unrelated resources. The minimum repayment instruction aims to maintain the account's existence and prevent complete default. The resource sourcing prompts are not simple text but are specific suggestions about potential resources or external resource sources generated after traversing the target resource account's related resource nodes using knowledge graph technology.

[0208] In step 30356, when the transaction rationality index indicates low rationality, a second type of combined resource repayment strategy is generated, which includes a minimum amount repayment instruction and consumption optimization prompts.

[0209] In practice, when the transaction rationality index indicates low rationality (e.g., below 0.5), regardless of the degree of stress, the predicament of the target resource account is considered to stem from poor resource consumption habits. In this case, the generated second type of combined resource return strategy includes a minimum return instruction and consumption optimization suggestions. Here, the consumption optimization suggestions are targeted structural adjustment recommendations generated after identifying non-productive high-frequency transaction tags (e.g., high-value entertainment, disorderly expansion investment) in the historical transaction data set based on natural language processing technology and cluster analysis algorithms.

[0210] As an example, if the system identifies that the target resource account has an excessively high proportion of gaming spending, it generates a spending optimization tip suggesting that the proportion of digital entertainment spending be reduced to below 10%.

[0211] Using the above methods, based on the transaction rationality index, liquidity buffer ratio, and installment return score, a second type of combined resource repayment strategy is generated through a preset decision matrix. First, the ratio of total installment service fee to installment fee is calculated to determine whether installment repayment has positive returns. Then, long-term, medium-to-long-term installment repayment instructions or over-repayment instructions are adapted according to different rationality and pressure levels. At the same time, minimum repayment instructions and corresponding prompts are provided for medium and low rationality scenarios. This can accurately adapt to the actual situation of the target resource account and improve the scientificity and pertinence of the resource repayment strategy.

[0212] In step 304, if the available resources are less than the minimum amount of resources to be returned, a combined resource return strategy including a resource overdue warning is generated.

[0213] In practice, resource overdue warnings serve as a high-priority risk notification signal, clearly alerting users that their current available resources will lead to overdue repayments. When available resources fall below the minimum repayment amount, a combined resource repayment strategy is immediately generated. This strategy not only includes the resource overdue warning but also a quantitative preview of the consequences of overdue payments, such as credit score deductions for the target account and increased additional repayment amounts. It also provides an entry point for emergency contact with customer service or application for an extension.

[0214] As an example, if the available resource quantity is only 100 and the minimum resource quantity to be returned is 500, a resource overdue warning message will be generated: "Overdue will result in an estimated additional resource return of 50. Please handle it immediately."

[0215] Using the above method, corresponding resource repayment strategies are generated when the available resources are in different ranges. First, credit risk is judged based on the normalized value of historical overdue payments. Then, a transaction rationality index is obtained by weighting the proportion of multiple types of transactions. The liquidity buffer rate is calculated by combining the resource difference and necessary expenditures, and then the installment income score is determined. Based on the decision matrix, appropriate installment or repayment instructions are output, and overdue warnings are output when resources are insufficient. This can comprehensively assess the account status, accurately control risks, and improve the rationality and reliability of resource repayment arrangements.

[0216] In some embodiments, after step 105, the combined resource return strategy is presented by the following method: generating a resource composition distribution view based on the resource quantity statistics under each level category; generating a pressure status dashboard component based on the resource pressure level; and returning the resource return feedback information carrying the resource composition distribution view and the pressure status dashboard component to the terminal that sent the resource return request, so as to display the resource composition distribution view and the pressure status dashboard component on the terminal.

[0217] In practical implementation, a resource composition distribution view is a visual carrier that maps multi-dimensional resource allocation data into a two-dimensional planar geometric figure, typically represented as a pie chart, Sankey diagram, or multi-layered ring diagram; this application does not impose any limitations on this. Generating a resource composition distribution view can be achieved through a geometric mapping algorithm. Specifically, it involves obtaining the statistical values ​​of resource quantities under each category (e.g., productive resource input, necessary survival consumption, and non-essential hedonistic consumption) calculated in the preceding steps, and then calculating the proportion coefficient of each category of resource quantity in the total resource quantity. Based on the proportion coefficient, an angle mapping formula is used... ,in, For the first The category proportion coefficient determines the central angle size of each category in the circular view, or the pixel bandwidth occupied by each category in the hierarchical flow diagram is determined using a coordinate mapping algorithm. A data object containing color coding and legend annotations is generated by the graphics rendering engine; this data object constitutes the resource composition distribution view. Furthermore, it can be extended to generate a dynamic, interactive 3D hierarchical view, constructing a 3D coordinate space where the main resource category is mapped to the base region, and subcategories are mapped to superimposed height vectors. Lighting rendering technology distinguishes the health levels of resources at different levels (e.g., high-health productive inputs are displayed as high-brightness areas), allowing users to view the deep structure of resource composition through rotation.

[0218] The process of generating a pressure status dashboard component based on resource pressure levels involves the conversion of scalar values ​​to vector graphics. Here, the pressure status dashboard component is a graphical user interface control that simulates a physical dashboard and is used to characterize the liquidity risk level of a target resource account in real time.

[0219] Specifically, the resource stress level calculated in the previous steps (usually a normalized floating-point number between 0 and 1) is read. Based on this, a semi-circular scale space (e.g., 0 to 180 degrees) is defined, and a color gradient band is set (e.g., transitioning from a green safe zone to a red danger zone). According to the formula... Calculate the deflection angle of the pointer ,in, This represents the resource pressure level. Furthermore, the pointer jitter parameter is calculated based on the rate of change of the pressure level to simulate the dynamic changes in real-time pressure. The component generation can also incorporate time series analysis technology, overlaying historical pressure fluctuation curves onto the dashboard background. Resource pressure level data from a preset time period (e.g., the past 12 periods) is extracted, and a trend background layer is generated using a Bézier curve smoothing algorithm. This ensures that the current pointer position not only reflects instantaneous pressure but also, combined with the background curve, demonstrates the evolution trend of pressure (e.g., whether it's in an upward or downward channel).

[0220] As an example, if the current resource stress level is 0.75 (high stress), the pointer of the dashboard component will point to the 135-degree position, and the background of the area will be rendered in dark red. At the same time, a high-risk warning text label will be displayed below the component.

[0221] In practical implementation, a two-way data communication link is established to return resource return feedback information, carrying the resource composition distribution view and pressure status dashboard components, to the terminal that sent the resource return request for display. Specifically, the data object of the aforementioned generated resource composition distribution view (such as a vector graphics description file) and the configuration parameters of the pressure status dashboard components (such as pointer angle and color threshold) are encapsulated in a standardized Hypertext Transfer Protocol (HTTP) response message. This response message, as resource return feedback information, is transmitted to the requesting user terminal via an encrypted channel. Upon receiving this feedback information, the application on the user terminal calls the local graphics rendering library to parse the data object and draw the view and components in a designated area of ​​the screen. The resource return feedback information also includes an interactive logic script. When the user clicks on a specific sector in the resource composition distribution view, the script triggers a zoom-in action to display the specific transaction details under that category. When the user long-presses the pressure status dashboard component, a predictive simulation function is triggered, dynamically displaying the pointer's fall position after executing the recommended return strategy.

[0222] Using the above methods, a resource composition distribution view is generated based on the statistical results of resource quantities at each level, and a pressure status dashboard component is generated according to the resource pressure level. Multi-dimensional and visualized data display is achieved through geometric mapping algorithms and graphic rendering. At the same time, time series analysis is combined to present the pressure evolution trend. Resource return feedback information carrying the view and components is returned to the terminal for display and interactive operation is supported. It can intuitively display the resource structure and pressure status, improve the user's understanding of the account status, and enhance the clarity and interactive experience of the resource return strategy display.

[0223] In some embodiments, the segmented repayment instruction includes at least one of the following: a minimum amount repayment instruction, a full amount repayment instruction, or an installment repayment instruction; after step 105, the segmented repayment instruction is also executed by: receiving a confirmation signal for the combined resource repayment strategy; in response to the confirmation signal, invoking the resource agency repayment interface to batch execute the minimum amount repayment instruction, full amount repayment instruction, or installment repayment instruction included in the segmented repayment instruction.

[0224] Here, segmented repayment instructions are defined as a set of operation commands targeting a specific resource account, with specific time and resource attributes. Specifically, the minimum repayment instruction refers to repaying only the minimum amount of resources required to maintain the account's credit record (e.g., repaying 10% of the outstanding resources); the full repayment instruction refers to repaying all outstanding resources for the current period in one lump sum; and the installment repayment instruction refers to splitting the outstanding resources for the current period according to preset periods and rates, generating repayment plans corresponding to multiple future time points. These three instruction types correspond to different levels of resource pressure relief and credit maintenance strategies.

[0225] In practice, the confirmation signal is a digitally signed data packet triggered through a specific interactive interface by the user terminal or authorized agent after reviewing and approving the recommended combined resource return strategy. The confirmation signal not only contains the user's consent identifier but also carries a security verification token to ensure the legality and non-repudiation of the instruction execution. This ensures that the automatically generated strategy has been subjectively confirmed by the user before execution, complying with the principle of informed consent in resource transactions. Furthermore, it can also require receiving biometric authentication signals from the user terminal and secondary verification signals from associated trusted devices (such as bound mobile tokens). Only when all confirmation signals pass the consistency check is a valid execution trigger instruction generated.

[0226] Upon receiving a valid confirmation signal, the system enters the execution phase, invoking the resource agency return interface. This interface is a standardized application programming interface connecting the strategy generation system and the core resource processing system, possessing high concurrency capabilities and transaction atomicity guarantees. An encrypted communication tunnel is established through the resource agency return interface, creating a secure command transmission channel. Based on this, the minimum amount return instruction, full amount return instruction, or installment return instruction contained within the segmented return instructions are executed in batches.

[0227] Here, batch execution refers to encapsulating multiple atomic return instructions into a single transactional request package and sending it to the resource organization's return interface all at once. This reduces network latency and ensures operational consistency. If any instruction in the request package fails to execute, the executed portion of the instructions will be rolled back according to a pre-defined transaction rollback mechanism, or an exception handling process will be triggered.

[0228] Furthermore, when invoking the resource agency's return interface, the optimal execution path can be dynamically selected based on the real-time load status and success rate statistics of each return channel. For installment return instructions, corresponding installment contracts are automatically generated and electronically signed and archived in the background.

[0229] By employing the above methods, after the combined resource return strategy is generated, a confirmation signal is received and security verification is completed to ensure that the execution of instructions is legal, compliant, and in accordance with the principle of informed consent. After responding to the confirmation signal, the resource agency's return interface is invoked to execute minimum amount return instructions, full amount return instructions, or installment return instructions in batches. This reduces network latency and ensures operational consistency. The transaction rollback mechanism and dynamic path selection enhance execution stability. At the same time, the installment contract archiving is automatically completed, improving the security, efficiency, and standardization of the execution of segmented return instructions.

[0230] The following will describe an exemplary application of the embodiments of this application in a real-world application scenario.

[0231] With the widespread adoption of electronic resource interaction methods, users face a variety of return options when returning resources to a target account. Different return methods correspond to different demands on available resources and additional service resource consumption (such as service fees and time costs). However, related technologies generally employ fixed return guidance schemes, failing to provide personalized suggestions based on the user's actual resource interaction structure and current available resource status. This can lead to liquidity depletion for resource-constrained users who choose full return, or excessive reliance on minimum returns resulting in high additional resource burdens. In actual business operations, unreasonable return decisions commonly result in decreased user experience, low fulfillment rates, and increased institutional bad debt risks. Therefore, in the field of resource return recommendation, there is an urgent need for a technological solution capable of intelligently identifying the importance of interactions, accurately assessing account resource pressure, and personalized matching of return strategies.

[0232] The relevant technology uses fixed rules for repayment guidance, typically presenting users with three standard options after the historical transaction data set is generated: full repayment, minimum repayment (usually a fixed percentage of the total amount), and installment repayment (fixed number of periods). The amount of resources to be repaid and additional costs corresponding to each option are simply calculated based on the total resource amount and credit level, and users are notified to make their selection. The entire guidance process is based on hard-coded rules across two dimensions: total resource amount and historical repayment records, without analyzing the content of specific transaction data units. The shortcomings of this approach are: it completely ignores the structural differences within the transaction data set, failing to distinguish the importance of essential interactions (such as survival-related interactions) from non-essential interactions (such as enjoyment-related interactions); the assessment of account resource status is too crude, relying solely on historical overdue records to fail to reflect the current true resource pressure; the fixed repayment options lack flexibility, failing to provide a "prioritize repayment of important interactions" combination solution when available resources are insufficient; and it lacks adaptive capabilities, unable to optimize recommendation strategies based on feedback.

[0233] Some related technologies introduce external scoring mechanisms to assist in repayment recommendations. By acquiring users' external credit scores and other data, and combining this with the total amount of resources, a repayment capability index is calculated. Users are then divided into different risk levels based on their repayment capability index, and a single repayment method is recommended accordingly (e.g., high-risk users are forced to repay the minimum amount). The drawbacks of this approach are that external scoring mainly reflects historical credit behavior and cannot reflect the resource interaction structure characteristics within the current period (e.g., it cannot distinguish whether resource outflow is for basic survival or for extended enjoyment); the judgment of resource pressure relies on external data sources, leading to data delays and compliance risks; the fixed strategy based on risk levels is too rigid and does not consider the differences in the deferability of transaction data units; and it lacks semantic understanding capabilities, making it impossible to extract scenario information from interaction summaries and identifiers.

[0234] Some related technologies also use predictive models to predict the resource return behavior of target resource accounts. By collecting features such as historical transaction data units and return records, the model is trained to predict the probability of users choosing different return methods and recommend the option with the highest probability. The shortcomings of this approach are that it is mainly trained based on historical behavioral features, lacking a deep understanding of the content of transaction data units, and the recommendation logic remains at the level of statistical correlation; it does not make sufficient use of semantic information, ignoring key textual information in transaction summaries; the model predicts the methods that users may choose, rather than the optimal options that users should choose, which can easily reinforce irrational decision-making habits; and the model has poor interpretability, making it difficult to explain the basis for the recommendation to users.

[0235] Therefore, the resource return processing solutions of related technologies have the following unresolved problems: Lack of interactive semantic understanding capabilities; related technologies generally ignore the structural differences within historical transaction datasets and cannot distinguish between survival-oriented and enjoyment-oriented categories through attribute features, resulting in return recommendations deviating from the actual needs of maintaining normal account operation; Inaccurate resource pressure assessment; related technologies rely solely on total resource volume or historical records, failing to quantify the account's current actual resource load and easily leading to misjudgments of account risk levels; Single and inflexible strategies; related technologies can only provide standardized single instructions and cannot generate combined resource return strategies that include segmented return instructions, making it difficult to achieve refined scheduling of prioritizing survival-oriented interactions and deferring enjoyment-oriented interactions when resources are limited; Lack of theoretical support and personalization in recommendation logic; the logic of related technologies cannot adapt to the individual differences of different accounts, lacks a quantitative analysis framework based on necessity scores and comprehensive scores, and cannot provide interpretable decision support from the perspective of resource allocation optimization.

[0236] To address the aforementioned issues, this application provides a resource return processing method. By introducing attribute feature analysis and semantic recognition mechanisms, a hierarchical classification system is established, including survival, security, development, and enjoyment categories. This system accurately identifies the importance of each transaction data unit based on necessity scores, extracts key information reflecting the true survival status of the account from historical transaction data sets, and provides a data foundation for strategy generation. A resource pressure assessment model is constructed, comprehensively calculating the proportion of survival-type transactions, periodic resource inflows, and historical overdue situations to quantify the resource pressure level of the target resource account. Starting from the essence of the resource interaction structure, it corrects the bias of single-dimensional assessment and improves the accuracy of identifying account liquidity risk. A multi-dimensional assessment model is used to determine the comprehensive score of each transaction data unit, and when available resources are insufficient, a combined resource return strategy including segmented return instructions (such as minimum amount return instructions combined with installment return instructions) is dynamically generated. This breaks through the limitations of traditional fixed options, achieving refined and atomic resource return operations, protecting the account's credit record while alleviating liquidity pressure caused by full return.

[0237] This application employs a dual identification mechanism of merchant category identifier and transaction summary information, combined with attribute feature weighted calculation, to accurately map transaction data units in the historical transaction dataset to multiple hierarchical categories, including survival-type categories. Simultaneously, by constructing a resource pressure assessment model, it comprehensively calculates the proportion of survival-type transactions, resource load ratio, and historical overdue data. The quantitatively determined resource pressure level objectively reflects the true resource turnover capacity of the target resource account. A comprehensive score is calculated by combining the necessity score, urgency weight, and resource quantity proportion of each transaction data unit, and a combined resource return strategy containing segmented return instructions is intelligently generated when available resources are limited. This differentiated strategy generation mechanism prioritizes the return of resources for high-necessity units and reasonably postpones the processing of low-priority units, achieving refined management and personalized optimization of resource allocation for the target resource account, significantly improving the rationality of resource scheduling and risk control capabilities.

[0238] In some embodiments, see Figure 13 , Figure 13 This is a schematic diagram illustrating the operation of the resource return processing method provided in this application embodiment. The resource return processing method provided in this application embodiment can be applied to related resource return applications. After the application is started, the application can automatically query and aggregate the historical transaction data units of the target resource account from the background query interface, and present them in a card-like format. Figure 13The resource return information display interface 1301 presents the data of the target resource account in a card-like format. Using the resource return processing method provided in this embodiment, the transaction data units are semantically parsed and hierarchically classified to identify four consumption levels: survival, security, development, and enjoyment, which are then labeled with different tags on the interface. Simultaneously, the user's resource pressure level is assessed, generating differentiated combined resource return strategies. These strategies, along with their advantages and reasons, are presented in the combined resource return strategy display interface 1302 for the user's reference and operation. In the resource return information display interface 1301, the user can also click on the card-like target resource account data to modify or enter data such as the current available resource quantity.

[0239] In some embodiments, see Figure 14 , Figure 14 This is a schematic diagram of the eleventh step in the resource return processing method provided in this application embodiment; through Figure 14 The multiple stages presented in this application's embodiment illustrate the overall processing flow. After the target object inquires about the resource return method instruction 1401, in response to the target object's resource return consultation request for the target resource account, the process enters the resource data acquisition and preprocessing stage 1402. The process retrieves the historical transaction data set within the target resource account's period from the data interface and performs data cleaning and formatting operations to transform the unstructured raw records into standardized transaction data units.

[0240] After the data is prepared, in the semantic recognition and hierarchical category recognition stage 1403, a dual recognition mechanism is used to perform in-depth analysis of each transaction data unit: on the one hand, preliminary classification mapping is performed based on the interactive object identifier (such as attribute code) (weight, for example, 0.6); on the other hand, natural language processing and semantic feature extraction are performed based on transaction summary information (weight, for example, 0.4). Through comprehensive weighted calculation, each transaction data unit is accurately divided into preset levels such as survival category, security category, development category, and enjoyment category. Next, in the resource pressure level assessment stage 1404, based on the classification results, multi-dimensional indicators such as the proportion of survival transactions, the ratio of total outflow of resources to periodic inflow of resources, and historical resource scheduling delay behavior are comprehensively calculated to quantitatively output the resource pressure level (or resource tension index) reflecting the current resource turnover load status of the account.

[0241] Based on this, in the transaction scoring and ranking stage 1405, for each transaction data unit, a comprehensive score is calculated by combining factors such as its necessity score, time urgency, resource volume proportion, and user historical interaction preferences, thereby generating an ordered sequence representing the priority of return. Subsequently, in the combined resource return strategy generation stage 1406, the optimal strategy is dynamically generated based on the comparison between the current available resources and the total amount of resources to be returned: when available resources are sufficient, a decision tree model is used to recommend a full return instruction, an intelligent segmented return instruction, or a minimum standard return instruction based on the resource pressure level; when available resources are limited but meet the minimum standard, a combined resource return strategy is intelligently generated based on multi-dimensional evaluation such as the interaction rationality index, liquidity buffer rate, and segmented execution benefit score, which includes prioritizing high-priority units and performing delayed or phased execution on low-priority units; when available resources are extremely scarce (not meeting the minimum standard), a resource scheduling delay risk warning is generated.

[0242] In step 14021 of the resource data acquisition and preprocessing stage 1402 described above, the historical transaction data set and its corresponding periodic summary data are acquired. The historical transaction data set and its corresponding periodic summary data of the target resource account are retrieved through a standard data interaction interface. The periodic summary data includes the account period identifier, total amount of resources to be returned, minimum standard return amount, return deadline, and resource scheduling delay (overdue) records. The historical transaction data set includes fields such as the interaction timestamp, resource outflow, interaction object name, interaction object category identifier, and transaction summary information for each transaction data unit. The available resource quantity is input by the target object through a human-computer interaction interface.

[0243] In step 14022, data preprocessing and field extraction are performed. The acquired raw data is cleaned and formatted, specifically including: identifying and removing invalid data records with attributes of resource rollback, operation reversal, or accounting adjustment based on the interaction type field; and filtering and retaining valid transaction data units (normal consumption) representing normal resource outflow. Then, the interaction timestamp field is uniformly converted to a standard time measurement format (such as UNIX timestamp or ISO 8601 standard format). For transaction data units lacking interaction object category identifiers, the interaction object name is used to search and match in a pre-set interaction object attribute database to complete the identifier information (completing the merchant classification code). Text normalization processing is performed on the transaction summary information, including removing special character interference and unifying the character encoding format (such as UTF-8) to ensure the accuracy of subsequent semantic parsing (preprocessing consumption description). Simultaneously, key field integrity checks are performed to confirm that at least two of the three core dimensions—resource outflow volume, interaction timestamp, and interaction object information (merchant information)—exist valid data to ensure the usability of the transaction data units.

[0244] In the aforementioned semantic recognition and hierarchical category recognition stage 1403, a dual-path fusion mechanism of interactive object category identifier mapping and transaction summary semantic parsing is adopted to perform necessity score calculation for each transaction data unit (each consumption) and determine its hierarchical category affiliation based on the score results.

[0245] In step 14031, based on the category identifier of the interactive object in the transaction data unit, a first basic score is determined by retrieving the built-in identifier mapping rule table. Based on the category identifier of the interactive object in the transaction data unit, the four-digit code is directly mapped to the corresponding first basic score (basic score) by retrieving the built-in identifier mapping rule table (merchant classification code and consumption level mapping rule table). This mapping table covers score definitions for different categories such as daily necessities supply (supermarkets / grocery stores), medical services (doctors / clinics), catering services (restaurants), and high-end consumer goods (jewelry), serving as the primary basis for judgment, and is assigned a first weight (e.g., 0.6).

[0246] For example, the contents of some merchant classification codes and consumption level mapping rule tables include {Merchant classification code: 4111; Merchant type: Public transportation: Basic score: 70, Merchant classification code: 4121; Merchant type: Taxi / Ride-hailing; Basic score: 62, Merchant classification code: 4814; Merchant type: Telecommunications services; Basic score: 68}.

[0247] In step 14032, text segmentation is performed on the transaction summary information in the transaction data unit to determine the second semantic score. Text segmentation is then performed on the transaction summary information (consumption description text) in the transaction data unit to extract key feature words. Subsequently, the extracted feature words are matched against a preset scenario keyword library. The keyword library is organized according to a preset hierarchy, including: survival maintenance (survival-related, such as hospital, rent, utilities), security (safety-related, such as insurance, school), development and improvement (development-related, such as phone bill, transportation), and enjoyment and experience (enjoyment-related, such as tourism). The number of hits of feature words in each category library is counted, the matching degree is calculated, and the corresponding second semantic score (semantic score) is output. This score is used to identify fine-grained scenarios, and a second weight (e.g., 0.4) is set.

[0248] In step 14033, a weighted fusion calculation of the necessity score is performed. The first basic score and the second semantic score are weighted and summed according to a set weight ratio to calculate the resource necessity score of the transaction data unit. The calculation formula is expressed as: Necessity score = First basic score × First weight + Second semantic score × Second weight.

[0249] In step 14034, consumption level classification is determined based on necessity scores. According to the calculated resource necessity scores, transaction data units are divided into corresponding preset level categories: when the necessity score is in the first preset range (e.g., 90 to 100 points), it is classified as a survival category (covering food, medical care, basic utilities, etc.); when the necessity score is in the second preset range (e.g., 70 to 90 points), it is classified as a security category (covering insurance, education, public transportation, etc.); when the necessity score is in the third preset range (e.g., 50 to 70 points), it is classified as a development category (covering communication, training, etc.); and when the necessity score is in the fourth preset range (e.g., 0 to 50 points), it is classified as an enjoyment category (covering entertainment, high-end dining, etc.). This process accurately identifies the category attributes of transaction data units such as "XX Hospital - Outpatient XX Fees," ensuring the accuracy of subsequent resource pressure assessments.

[0250] For example, the transaction summary information in a certain transaction data unit is "XX Hospital - Outpatient XX Fees", merchant classification code 8011, merchant classification code mapping score 95 points, text parsing extraction of "hospital" and "outpatient" keywords score 98 points, and the fusion calculation necessity score = 95 × 0.6 + 98 × 0.4 = 96.2 points. Based on the necessity score, the consumption level is divided and determined to be subsistence.

[0251] In the resource stress level assessment phase 1404, the current load status of the target resource account is calculated based on three dimensions of quantitative indicators, and a normalized resource stress index is output, ranging from 0 to 1. In step 14041, the proportion of survival-type resource outflow is calculated. Based on the resource allocation principle in economics (Engel's coefficient principle), the proportion of survival-type resource outflow is calculated. This indicator is obtained by dividing the total amount of transaction data units marked as survival-type by the total amount of resources to be returned in the current period, and is used to characterize the degree to which basic survival needs occupy available resources.

[0252] Then, the second dimension indicator is calculated, namely, the periodic resource outflow to inflow ratio in step 14042. This indicator is obtained by dividing the total amount of resources to be returned in the current period by the periodic resource inflow of the target object. The resource inflow data can be obtained through a data interface or manually entered by the user, and is used to reflect the resource return pressure.

[0253] Next, the third dimension indicator is calculated, namely, the normalized value of resource scheduling delay in step 14043. The normalized value of resource scheduling delay corresponds to the normalized value of the number of overdue events mentioned above. This indicator is obtained by statistically analyzing the number of resource scheduling delay records within a preset historical period (overdue events in the past 6 months) and normalizing them (for example, dividing by the number of statistical periods, 6), and is used to reflect historical resource management capabilities and credit risk.

[0254] In step 14044, the resource stress index is calculated and determined. A weighted index is calculated by summing the three indicators based on preset weighting coefficients (e.g., survival rate 0.5, outflow-to-inflow ratio 0.3, and normalized value of overdue number 0.2), outputting a quantified resource stress index. The weighted calculation formula can be expressed as: Resource Stress Index = First Dimension Indicator × 0.5 + Second Dimension Indicator × 0.3 + Third Dimension Indicator × 0.2.

[0255] After calculating and determining the resource stress index, step 14045 is executed to classify stress levels based on the stress index. Users are divided into different resource stress levels according to the calculated resource stress index: when the resource stress index is greater than the first stress threshold (e.g., >0.7), it is determined to be a high stress level (high stress user); when the resource stress index is between the first and second stress thresholds (e.g., between 0.4 and 0.7), it is determined to be a medium stress level (medium stress user); when the resource stress index is less than the second stress threshold (e.g., <0.4), it is determined to be a low stress level (low stress user).

[0256] For example, if the total amount of resources to be repaid is 8,000 yuan, of which 3,500 yuan is survival-related resources outflow and 10,000 yuan is resources inflow, and there is one overdue payment in the past 6 months, the financial stress index is calculated as follows: (3,500 / 8,000)×0.5+(8,000 / 10,000)×0.3+(1 / 6)×0.2=0.219+0.24+0.033=0.492, which is considered a medium stress user.

[0257] In the aforementioned transaction scoring and ranking stage 1405, a comprehensive score is calculated for each transaction data unit within the period based on multi-dimensional evaluation indicators, and a priority sequence is generated accordingly to provide a basis for subsequent resource scheduling strategies. The comprehensive evaluation dimensions cover four aspects: necessity attributes, time urgency, resource occupancy ratio, and historical behavior preferences. The first dimension indicator is calculated, i.e., step 14051, to calculate the necessity weight. This involves determining the necessity weight by directly referencing the resource necessity score calculated in the previous stage and converting it into a weight value through normalization. This weight value characterizes the importance of the transaction data unit in terms of survival and maintenance, and its weight coefficient is set to a first preset value (e.g., 0.4).

[0258] Next, calculate the second dimension indicator, i.e., execute step 14052 to calculate the urgency weight. That is, determine the urgency weight of the time dimension. This indicator is calculated based on the ratio of the interval between the transaction time and the current time point to the account cycle. The earlier the interaction timestamp, the longer its existence within the billing cycle, and the higher the urgency of repayment. It can be calculated by the company ((current date - consumption date) / billing cycle days) and its weight coefficient is set to the second preset value (e.g., 0.3).

[0259] Next, the third dimension indicator is calculated, namely, step 14053, which calculates the percentage of a single transaction's consumption amount. This determines the resource volume percentage, which is calculated by the ratio of the resource outflow of a single transaction data unit to the total amount of resources to be returned in the current period. It is used to reflect the resource load of a single transaction on the overall resource account, and its weighting coefficient is set to a third preset value (e.g., 0.2).

[0260] Then, the fourth dimension indicator is calculated, i.e., step 14054, which calculates preferences based on historical repayment behavior. This involves determining the historical behavior preference coefficient. This indicator is obtained by analyzing the target object's resource scheduling records over a historical period and statistically analyzing the frequency with which different levels and categories of transactions are prioritized. If the target object lacks historical records (new users), then preset default parameters (default preference values, such as 0.9 for survival type, 0.8 for security type, etc.) are used; and its weight coefficient is set to the fourth preset value (e.g., 0.1).

[0261] Finally, step 14055 is executed to calculate the weighted overall consumption importance score. A comprehensive score and ranking are performed. Based on the calculation results of the above four dimensions and their corresponding weight coefficients, a weighted sum is calculated to obtain the comprehensive score for each transaction data unit. All transaction data units are then arranged in descending order of their comprehensive scores, generating an ordered resource scheduling sequence (corresponding to the above transaction data unit sequence).

[0262] For example, a transaction data unit is 2000, with a necessity score of 96.2, a consumption date 25 days from the resource return date (billing cycle 30 days), a total resource to be returned of 8000 yuan, and a user's historical priority consumption frequency of 0.9. In this case, the comprehensive score is calculated as follows: (96.2 / 100)×0.4+(25 / 30)×0.3+(2000 / 8000)×0.2+0.9×0.1=0.385+0.250+0.05+0.09=0.775.

[0263] In the aforementioned combined resource return strategy generation stage 1406, step 14061 is executed to generate a combined resource return strategy. Based on the consumption level classification results, resource importance ranking list, resource pressure level, and the current available resource status of the target object generated in the previous stage, a differentiated resource return plan (personalized return plan) is dynamically generated. The combined resource return strategy generation stage 1406 generates strategies in three scenarios based on the matching relationship between the current total available resources, the total amount of resources to be returned, and the minimum amount of resources to be returned.

[0264] In the first scenario, when it is determined that the total amount of currently available resources is greater than or equal to the total amount of resources to be returned, the decision tree logic is executed based on the resource pressure level (funds tightness index) and the consumption hierarchy.

[0265] In some embodiments, see Figure 15 , Figure 15 This is the first example diagram generated by the strategy provided in the embodiment of this application; when the total amount of available resources is greater than or equal to the total amount of resources to be returned, the capital tension index is judged.

[0266] When the funding stress index > 0.7 (corresponding to the first stress threshold mentioned above), the target resource account is considered a high-stress user. Step 1502 is executed to generate a first-type combined resource repayment strategy based on the minimum resource repayment amount. That is, if the resource stress level is high stress level, a first-type combined resource repayment strategy based on the minimum resource repayment amount is generated. Step 1503 is further executed to determine whether there is a specific large-value transaction data unit. If the determination result is yes, a single installment plan is recommended, and an independent resource segmentation scheduling scheme (corresponding to the single installment repayment instruction of the target transaction data unit) is recommended for the specific large-value transaction data unit marked as highly necessary (e.g., single transaction > 1000 yuan and score > 90). If the determination result is no, only the minimum resource repayment amount is recommended.

[0267] When the funding stress index is within the closed range of 0.4 (corresponding to the second stress threshold mentioned above) to 0.7, i.e. (0.4 ≤ funding stress index ≤ 0.7), the target resource account is considered a medium-stress user, and step 1504 is executed to recommend a global resource installment repayment strategy. That is, if the resource stress level is medium-stress, a global resource installment repayment strategy is recommended (corresponding to the first type of combined resource repayment strategy including installment repayment instructions mentioned above). Step 1505 is then executed to determine the segmented period comparison results of the proportion of enjoyment-type consumption. The segmented period is dynamically adjusted based on the proportion of non-essential resources (the proportion of enjoyment-type consumption). For example, if the comparison result indicates a proportion > 40% (corresponding to the first proportion threshold mentioned above), a 6-month repayment plan + non-essential consumption reminder is recommended; if the comparison result indicates a proportion < 40% (corresponding to the second proportion threshold mentioned above) < 40%, a 3-month installment plan + non-essential consumption reminder is recommended; if the comparison result indicates a proportion ≤ 30%, a 3-month installment plan is recommended.

[0268] When the funding stress index is less than 0.4, the target resource account is considered a low-stress user, and step 1506 is executed, recommending a full resource return plan. That is, if the resource stress level is low, a full resource return plan is recommended, and a consumption level distribution + resource status good prompt is displayed (corresponding to the first type of combined resource return strategy that includes a full return instruction mentioned above).

[0269] In the second scenario, when the total amount of available resources is determined to be between the minimum amount of resources to be repaid and the total amount of resources to be repaid, a multi-dimensional assessment algorithm is executed to balance credit risk and liquidity pressure.

[0270] In some embodiments, see Figure 16 , Figure 16 This is a second example diagram generated by the strategy provided in the embodiments of this application; in step 1601, when it is determined that the total amount of currently available resources is between the minimum amount of resources to be returned and the total amount of resources to be returned, it is determined whether the normalized value of resource scheduling delay exceeds the preset risk threshold.

[0271] Here, the resource scheduling delay normalization value (overdue number normalization value) is checked. If it exceeds the preset risk threshold (e.g., ≥0.5), the forced locking strategy is to maximize repayment (repay as much as possible) to prevent credit deterioration (the reason being to maintain credit record). This corresponds to the above situation where the historical overdue normalization value of the target resource account within the preset statistical period is greater than or equal to the preset risk threshold, and a second type of combined resource repayment strategy containing a full repayment instruction is generated.

[0272] If the threshold is not exceeded, proceed to step 1602 to calculate the transaction rationality index. Based on the consumption level classification results, calculate the resource allocation rationality index (corresponding to the aforementioned transaction rationality index). The calculation logic is as follows: the weighted value of the proportion of survival and maintenance (survival consumption proportion × 1.0) plus the weighted value of the proportion of security and safety (security consumption proportion × 0.8) minus the weighted value of the proportion of enjoyment and experience (enjoyment consumption proportion × 0.5). The transaction rationality index reflects the structural characteristics of essential and non-essential consumption. An index ≥ 0.6 indicates high rationality (primarily necessities), 0.3~0.6 indicates medium rationality (balanced consumption), and < 0.3 indicates low rationality (excessive non-essential consumption).

[0273] Further execute step 1603 to calculate the liquidity buffer ratio. The formula for calculating the liquidity buffer ratio is: (Current available resources - Minimum repaid resources) / Forecast of short-term necessary resource expenditures (e.g., necessary expenditures in the next 7 days). A liquidity buffer ratio <1.0 indicates high liquidity pressure (severe shortage of subsequent available resources after the minimum repaid resources are repaid), 1.0~2.0 indicates medium pressure, and ≥2.0 indicates low pressure.

[0274] Then, proceed to step 1604 to calculate the phased revenue score. This involves calculating the value metric for segmented resource scheduling, namely the phased revenue score. The calculation combines resource allocation rationality indicators and liquidity redundancy indicators (e.g., transaction rationality index × 0.7 + (1 - liquidity buffer rate / 3) × 0.3), and compares it with the resource scheduling service fee rate (fee ratio).

[0275] Using the rationality index, liquidity index and benefit score calculated above, construct a multidimensional decision matrix, execute step 1605, determine the strategy matrix, and output the following combined resource return strategy.

[0276] For example, when the calculated resource allocation rationality index is higher than the preset high threshold and the liquidity redundancy index is lower than the preset safety bottom line (consumption rationality ≥ 0.6 and liquidity buffer rate < 1.0), a long-term global resource segmented scheduling scheme is generated (12-phase scheduling is recommended). This strategy aims to ensure liquidity buffer space by extending the scheduling cycle when the proportion of survival-maintenance resources is high, preventing the basic quality of life of the target object from being affected by excessive repayment pressure (the reason is that the proportion of survival-type consumption is high, and phased scheduling ensures the buffer of liquidity resources to avoid affecting the quality of life). This corresponds to the second type of combined resource repayment strategy that includes excess repayment instructions when the transaction rationality index indicates high rationality, the liquidity buffer rate indicates low pressure, and the phased scheduling has positive returns.

[0277] When the resource allocation rationality index is higher than the preset high threshold, the liquidity redundancy index is within the preset buffer range, and the resource scheduling value metric is higher than the service fee rate (consumption rationality ≥ 0.6, buffer rate between 1.0 and 2.0, and revenue score > fee ratio), a medium-to-long-term global resource segmented scheduling scheme is generated (6-12 periods are recommended). This strategy is based on a balanced resource structure and aims to control scheduling costs while maintaining necessary liquidity levels (the rationale being that the consumption structure is reasonable, and phased repayment can ensure liquidity while controlling costs). Corresponding to the above-mentioned second type of combined resource repayment strategy, which includes medium-to-long-term phased repayment instructions, under the conditions that the transaction rationality index indicates high rationality, the liquidity buffer rate indicates moderate pressure, and phased repayment has positive returns.

[0278] When the resource allocation rationality index is higher than the preset high threshold, the liquidity redundancy index is higher than the preset sufficiency threshold, and the resource scheduling value metric is higher than the service fee rate (consumption rationality ≥ 0.6, buffer rate ≥ 2.0, and return score > fee ratio), an excess resource repayment plan is generated (it is recommended to try repaying more). This strategy utilizes ample liquidity resources and suggests increasing the repayment amount above the minimum repayment threshold to reduce the interest cost of long-term holding (the reason being that liquidity resources are ample, and it is recommended to repay more to save costs). This corresponds to the second type of combined resource repayment strategy that includes excess repayment instructions when the transaction rationality index indicates high rationality, the liquidity buffer rate indicates low pressure, and the installment has positive returns.

[0279] When the resource allocation rationality index is in the preset intermediate transition range, and the liquidity redundancy index is below the preset safety threshold (consumption rationality between 0.3 and 0.6 and buffer rate < 1.0), a minimum repayment threshold scheme is generated with an additional external resource raising prompt (recommended minimum repayment + short-term resource raising). This strategy aims to address the current liquidity shortage and recommends retaining existing available resources and initiating remedial measures (the reason being the current liquidity crunch, suggesting retaining existing resources and raising additional resources as soon as possible for repayment). This corresponds to the second type of combined resource repayment strategy, which includes a minimum repayment instruction and a fund raising prompt, when the transaction rationality index indicates moderate rationality and the liquidity buffer rate indicates high pressure.

[0280] When the resource allocation rationality index is within a preset intermediate transition range, the liquidity redundancy index meets basic requirements, but the resource scheduling value metric is lower than the service fee rate (consumption rationality between 0.3 and 0.6, buffer rate ≥ 1.0, and revenue score < fee ratio), a standard minimum repayment threshold scheme is generated (minimum repayment is recommended). This strategy, after balancing liquidity needs and scheduling costs, prioritizes maintaining the flexibility of available resource scheduling (the rationale being that considering both liquidity needs and costs, maintaining flexibility is recommended). This corresponds to the second type of combined resource repayment strategy mentioned above, which includes a minimum repayment instruction and consumption optimization prompts when the transaction rationality index indicates low rationality.

[0281] When the rationality of consumption is less than 0.3, it is recommended to repay the minimum amount plus consumption optimization tips. The reason is that the proportion of enjoyment-type consumption is too high, and it is recommended to optimize consumption habits to avoid long-term installment costs.

[0282] Then, proceed to step 1606 to check if there are any large essential consumptions in the importance ranking list. Check the resource importance ranking list; if the check result is yes, a large essential transaction meeting specific conditions (essence score > 0.7 and high resource proportion) is found. Add a single-item scheduling suggestion (apply for single-item installment payment) to the main strategy. Add the suggestion "You have XX large essential consumptions; you can apply for single-item installment payment to reduce resource repayment pressure, providing users with more flexible resource repayment options" to the combined resource repayment strategy. If the check result is no, proceed to step 1607 to output the final recommended strategy.

[0283] In the third scenario, when the total amount of available resources is determined to be lower than the minimum return threshold, a high-risk warning signal is directly generated (indicating the risk of overdue payment) and a resource replenishment request is output.

[0284] For example, for a specific target object, the current period's resource status data is as follows: the total amount of resources to be returned is 7100, the minimum return threshold is 710, the current available total resources are 4000, and resource consumption hierarchy analysis shows: survival and maintenance resources account for 45% (survival-related consumption), safety and security resources account for 25% (safety-related consumption), and enjoyment and experience resources account for 9% (enjoyment-related consumption). Historical behavior records show that the number of resource scheduling delays in the past 6 periods is 0 (0 overdue times in the past 6 months), and the predicted value of short-term necessary resource expenditure is 1500 units (necessary expenditure of 1500 yuan in the next 7 days).

[0285] The normalized value of resource scheduling delay is calculated as 0 / 6 = 0. This value is less than the preset risk threshold (e.g., 0.5), indicating that the credit risk is controllable and the process proceeds to the multi-dimensional assessment. Based on the consumption level weight, the resource allocation rationality index (transaction rationality index) is calculated as 0.45×1.0+0.25×0.8-0.09×0.5=0.605. This result is higher than the high threshold (0.6), indicating a high rationality structure. The liquidity redundancy index (liquidity buffer ratio) is calculated as (4000-710) / 1500=2.19. This result is higher than the threshold (2.0), indicating a low liquidity pressure state. The resource scheduling value measure (installment income score) is calculated as 0.605×0.7+(1-2.19 / 3)×0.3=0.4235+0.081=0.504. Assuming the resource scheduling service fee rate (cost ratio) is calculated to be 8.6% (0.086), the comparison results show that the value measure (0.504) is significantly higher than the service fee cost (0.086), indicating that the scheduling revenue is greater than the cost.

[0286] Based on the decision matrix judgment logic, the target object meets the conditions of "resource allocation rationality index ≥ 0.6, liquidity redundancy index ≥ 2.0, and resource scheduling value measurement > service fee rate", and matches the strategy branch (high rationality, high liquidity and high return scenario) to generate excess resource return plan (it is recommended to try to return a part more).

[0287] The specific strategy recommendation is to increase the repayment amount by 1500 on top of the minimum repayment threshold. This aims to utilize ample liquidity resources to reduce long-term interest costs without impacting short-term survival quality. Finally, a scan of the resource importance ranking list revealed no large, essential transaction data units that met the single-transaction segmented scheduling criteria; therefore, no additional recommendations were generated. The final output is a complete resource repayment strategy including the rationale for the recommendation, specific amount arrangements, and financial analysis data.

[0288] Finally, after executing step 14061, the generated strategy is output in a visual form through the strategy presentation stage 1407. The displayed content includes an explanation of the recommendation reasons, a detailed list of resource allocation arrangements, and a financial analysis report based on the current resource composition distribution view, thereby assisting users in making rational resource scheduling decisions.

[0289] The resource return processing method provided in this application has the following beneficial effects: It improves the hierarchical classification accuracy of resource data units to be returned. By integrating the dual verification logic of resource provider identification and transaction semantic description, it effectively overcomes the limitations of single-dimensional identification and solves the semantic ambiguity problem in different application scenarios under the same resource provider. It enhances the accuracy of assessing the current available resource pressure status of the target object. The pressure assessment model based on the resource configuration structure can more accurately identify the pressure differences under different resource outflow and inflow patterns, especially for complex scenarios with high resource outflow but low inflow or low outflow but high inflow, effectively avoiding strategy generation deviations caused by misjudgment. Through dynamically generated differentiated resource return schemes, it effectively reduces the probability of resource scheduling delays for resource-scarce objects, while optimizing the resource scheduling cost structure for medium-pressure objects. This enables the target object to minimize resource scheduling costs while meeting liquidity needs, significantly improving the target object's satisfaction. Simultaneously, it enhances the ability to identify potential risks, enabling precise identification of potentially high-risk targets based on a comprehensive analysis of the proportion of survival-maintenance resources and resource pressure levels. This helps optimize the overall resource allocation strategy, thereby significantly reducing bad debt losses caused by inappropriate strategies. It possesses efficient real-time data processing capabilities, utilizing optimized semantic analysis algorithms to ensure rapid analysis and strategy generation for individual resource cycles. This meets the real-time response speed requirements of financial-grade systems and effectively supports concurrent access to a large number of target objects.

[0290] The following description continues to illustrate the exemplary structure of the resource return processing device 455 provided in the embodiments of this application as a software module. In some embodiments, such as Figure 2As shown, the software modules stored in the resource return processing device 455 in the memory 450 may include: a data acquisition module 4551, used to acquire the historical transaction data set and available resource quantity of the target resource account in response to a resource return request for the target resource account, wherein the historical transaction data set includes multiple transaction data units; a hierarchical category determination module 4552, used to determine the necessity score of each transaction data unit based on the attribute characteristics of each transaction data unit, and map multiple transaction data units to multiple hierarchical categories based on the necessity score, wherein the hierarchical categories include at least the survival category; a resource pressure level determination module 4553, used to determine the survival transaction ratio based on the transaction data units belonging to the survival category, and determine the resource pressure level of the target resource account based on the survival transaction ratio; a comprehensive score determination module 4554, used to determine the comprehensive score of each transaction data unit based on the necessity score and transaction attributes of each transaction data unit; and a strategy generation module 4555, used to generate a combined resource return strategy containing segmented return instructions based on the resource pressure level and the comprehensive score of each transaction data unit when the available resource quantity is less than the total resource quantity of the historical transaction data set.

[0291] In some embodiments, the attribute features include merchant category identifiers and transaction summary information; the hierarchical category determination module 4552 is further configured to match merchant category identifiers based on a preset identifier mapping table for each transaction data unit in the historical transaction data set, and determine the preset score corresponding to the matched merchant category identifier as the basic score of the transaction data unit; extract target keywords from the transaction summary information, and determine the semantic score of the transaction data unit based on the matching degree between the target keywords and each hierarchical category; and perform a weighted summation of the basic score and the semantic score according to a preset first weight ratio to obtain the necessity score of the transaction data unit.

[0292] In some embodiments, the hierarchical category determination module 4552 is further configured to obtain a preset hierarchical keyword library, the hierarchical keyword library including a set of reference keywords corresponding to each hierarchical category; for each hierarchical category, to count the number of hits of the target keyword in the reference keyword set; to determine the matching degree between the target keyword and the hierarchical category by the ratio of the number of hits to the total number of reference keywords in the reference keyword set; to select the target hierarchical category with the highest matching degree among multiple hierarchical categories, and to determine the preset score corresponding to the target hierarchical category as the semantic score of the transaction data unit.

[0293] In some embodiments, the hierarchical category determination module 4552 is further configured to obtain a preset category division threshold set, the category division threshold set including at least a first boundary threshold for defining the survival category; for each transaction data unit in the historical transaction data set, if the necessity score of the transaction data unit is greater than or equal to the first boundary threshold, the transaction data unit is mapped to the survival category.

[0294] In some embodiments, the category classification threshold set further includes a second boundary threshold and a third boundary threshold, wherein the first boundary threshold is greater than the second boundary threshold, and the second boundary threshold is greater than the third boundary threshold; the hierarchical category determination module 4552 is further configured to, for each transaction data unit in the historical transaction data set, map the transaction data unit to the security category if the necessity score of the transaction data unit is less than the first boundary threshold and greater than or equal to the second boundary threshold; map the transaction data unit to the development category if the necessity score of the transaction data unit is less than the second boundary threshold and greater than or equal to the third boundary threshold; and map the transaction data unit to the enjoyment category if the necessity score of the transaction data unit is less than the third boundary threshold.

[0295] In some embodiments, the resource pressure level determination module 4553 is further configured to sum the resource amounts of transaction data units belonging to the survival category to obtain a first total resource amount corresponding to the survival category; sum the resource amounts of each transaction data unit in the historical transaction data set to obtain a second total resource amount corresponding to the historical transaction data set; and determine the ratio of the first total resource amount and the second total resource amount as the survival transaction proportion.

[0296] In some embodiments, the resource stress level determination module 4553 is further configured to obtain the periodic resource inflow of the target resource account and the number of historical overdue transactions within a preset statistical period; determine the ratio of the second total resource amount to the periodic resource inflow as the resource load ratio of the target resource account; determine the ratio of the number of historical overdue transactions to the preset statistical period as the normalized value of the number of historical overdue transactions corresponding to the target resource account; calculate the resource stress index of the target resource account by weighting the survival transaction ratio, the resource load ratio, and the normalized value of the number of historical overdue transactions according to a preset second weighting ratio; and determine the resource stress level of the target resource account based on the resource stress index.

[0297] In some embodiments, the resource stress level determination module 4553 is further configured to obtain a preset first stress threshold and a preset second stress threshold; if the resource stress index is greater than the first stress threshold, determine the resource stress level of the target resource account as a high stress level; if the resource stress index is less than or equal to the first stress threshold and the resource stress index is greater than or equal to the second stress threshold, determine the resource stress level of the target resource account as a medium stress level; if the resource stress index is less than the second stress threshold, determine the resource stress level of the target resource account as a low stress level.

[0298] In some embodiments, the comprehensive score determination module 4554 is further configured to: normalize the necessity score of each transaction data unit to obtain the necessity weight of the transaction data unit; determine the urgency weight of the transaction data unit based on the transaction generation time and settlement reference time of the transaction data unit; determine the resource quantity ratio of the transaction data unit to the total resource quantity of the historical transaction data set as the resource quantity ratio of the transaction data unit; determine the preference coefficient corresponding to the transaction data unit based on the historical resource return behavior of the target resource account; and perform a weighted summation of the necessity weight, urgency weight, resource quantity ratio, and preference coefficient according to a preset third weight allocation to obtain the comprehensive score of the transaction data unit.

[0299] In some embodiments, the comprehensive score determination module 4554 is further configured to determine the time difference between the settlement reference time and the transaction generation time of the transaction data unit; and to determine the ratio of the time difference to the settlement cycle duration corresponding to the transaction data unit as the urgency weight of the transaction data unit.

[0300] In some embodiments, the comprehensive score determination module 4554 is further configured to detect whether there are valid historical resource return records in the target resource account and obtain a detection result; if the detection result indicates that there are historical resource return records in the target resource account, the frequency of transactions belonging to the hierarchical category to which the transaction data unit belongs is statistically analyzed and used as the preference coefficient corresponding to the transaction data unit; if the detection result indicates that there are no historical resource return records in the target resource account, the corresponding default preference value is matched from the preset cold start preference mapping table according to the hierarchical category to which the transaction data unit belongs and used as the preference coefficient corresponding to the transaction data unit.

[0301] In some embodiments, the strategy generation module 4555 is further configured to sort multiple transaction data units in the historical transaction data set based on the comprehensive score of each transaction data unit to obtain a transaction data unit sequence; and generate a combined resource repayment strategy containing segmented repayment instructions based on the resource pressure level and the transaction data unit sequence; wherein the segmented repayment instructions include at least one of the following: minimum amount repayment instruction, full amount repayment instruction, and installment repayment instruction.

[0302] In some embodiments, the strategy generation module 4555 is further configured to determine the total resource amount and the minimum resource amount to be returned corresponding to the historical transaction data set; when the available resource amount is greater than or equal to the total resource amount, generate a first type of combined resource return strategy based on the resource pressure level and the transaction data unit sequence; when the available resource amount is less than the total resource amount but greater than or equal to the minimum resource return amount, generate a second type of combined resource return strategy based on a multi-dimensional evaluation model; and when the available resource amount is less than the minimum resource return amount, generate a combined resource return strategy that includes a resource overdue warning.

[0303] In some embodiments, the strategy generation module 4555 is further configured to generate a first type of combined resource repayment strategy including a minimum repayment instruction when the resource pressure level is high; wherein, when there is a target transaction data unit in the transaction data unit sequence with a resource amount greater than a preset resource amount and a necessity score greater than a preset score threshold, the first type of combined resource repayment strategy includes a single installment repayment instruction for the target transaction data unit; when the resource pressure level is medium, a first type of combined resource repayment strategy including an installment repayment instruction is generated; wherein, when the resource proportion of transaction data units belonging to the enjoyment category in the transaction data unit sequence is greater than a first proportion threshold, the installment repayment instruction corresponds to a first installment period; when the resource proportion is between a second proportion threshold and a first proportion threshold, the installment repayment instruction corresponds to a second installment period; and when the resource pressure level is low, a first type of combined resource repayment strategy including a full repayment instruction is generated.

[0304] In some embodiments, the hierarchical categories further include: a security category, a development category, and an enjoyment category; the strategy generation module 4555 is further configured to generate a second type of combined resource repayment strategy containing a full repayment instruction when the historical overdue normalized value of the target resource account within a preset statistical period is greater than or equal to a preset risk threshold; when the historical overdue normalized value is less than the preset risk threshold, the proportion of survival-type transactions corresponding to the historical transaction data set, the proportion of security-type transactions corresponding to the security category, the proportion of development-type transactions corresponding to the development category, and the proportion of enjoyment-type transactions corresponding to the enjoyment category are weighted according to a preset fourth weight ratio. The transaction rationality index of the target resource account is obtained by summing the weights. The resource difference between the available resource quantity and the minimum repaid resource quantity corresponding to the historical transaction data set is determined, and the ratio of the resource difference to the necessary resource expenditure of the target resource account in the preset period is determined as the liquidity buffer ratio of the target resource account. Based on the transaction rationality index and the liquidity buffer ratio, the periodic return score of the target resource account is determined, wherein the periodic return score is positively correlated with the transaction rationality index and negatively correlated with the liquidity buffer ratio. Based on the transaction rationality index, the liquidity buffer ratio, and the periodic return score, a second type of combined resource repayment strategy is generated using a preset decision matrix.

[0305] In some embodiments, the strategy generation module 4555 is further configured to determine the total installment service fee based on the total resource amount corresponding to the historical transaction data set, the preset unit period service fee rate, and the preset number of installment periods, and to determine the ratio of the total installment service fee to the total resource amount as the installment fee ratio. Where the installment return score is greater than the installment fee ratio, it indicates that the installment repayment has a positive return. Where the transaction rationality index indicates high rationality and the liquidity buffer ratio indicates high pressure, a second type of combined resource repayment strategy containing long-term installment repayment instructions is generated. Where the transaction rationality index indicates high rationality, the liquidity buffer ratio indicates medium pressure, and the installment repayment has a positive return... Generate a second type of combined resource repayment strategy that includes medium- to long-term installment repayment instructions; when the transaction rationality index indicates high rationality, the liquidity buffer ratio indicates low pressure, and installment payments have positive returns, generate a second type of combined resource repayment strategy that includes excess repayment instructions, where the repayment amount indicated by the excess repayment instructions is higher than the minimum repayment amount; when the transaction rationality index indicates medium rationality, and the liquidity buffer ratio indicates high pressure, generate a second type of combined resource repayment strategy that includes minimum repayment instructions and funding prompts; when the transaction rationality index indicates low rationality, generate a second type of combined resource repayment strategy that includes minimum repayment instructions and consumption optimization prompts.

[0306] In some embodiments, after generating a combined resource return strategy containing segmented return instructions, the strategy generation module 4555 is further configured to generate a resource composition distribution view based on the resource quantity statistics results under each level category; generate a pressure status dashboard component based on the resource pressure level; and return the resource return feedback information carrying the resource composition distribution view and the pressure status dashboard component to the terminal that sent the resource return request, so as to display the resource composition distribution view and the pressure status dashboard component on the terminal.

[0307] In some embodiments, the segmented repayment instruction includes at least one of the following: a minimum repayment instruction, a full repayment instruction, and an installment repayment instruction; after generating a combined resource repayment strategy containing the segmented repayment instructions, the strategy generation module 4555 is further configured to receive a confirmation signal for the combined resource repayment strategy; in response to the confirmation signal, the resource organization repayment interface is invoked to batch execute the minimum repayment instruction, full repayment instruction, or installment repayment instruction contained in the segmented repayment instructions.

[0308] This application provides a computer program product, which includes a computer program or computer-executable instructions stored in a computer-readable storage medium. A processor of an electronic device reads the computer-executable instructions from the computer-readable storage medium and executes the computer-executable instructions, causing the electronic device to perform the resource return processing method described above in this application.

[0309] This application provides a computer-readable storage medium storing computer-executable instructions or a computer program. When the computer-executable instructions or the computer program are executed by a processor, the processor will execute the resource return processing method provided in this application. For example, ... Figure 3 The resource return processing method is shown.

[0310] In some embodiments, the computer-readable storage medium may be a memory such as RAM, ROM, flash memory, magnetic surface memory, optical disk, or CD-ROM; or it may be a variety of devices including one or any combination of the above-mentioned memories.

[0311] In some embodiments, computer-executable instructions may take the form of programs, software, software modules, scripts, or code, written in any form of programming language (including compiled or interpreted languages, or declarative or procedural languages), and may be deployed in any form, including as stand-alone programs or as modules, components, subroutines, or other units suitable for use in a computing environment.

[0312] As an example, computer-executable instructions may, but do not necessarily, correspond to files in a file system. They may be stored as part of a file that holds other programs or data, for example, in one or more scripts in a Hyper Text Markup Language (HTML) document, in a single file dedicated to the program in question, or in multiple co-located files (e.g., files that store one or more modules, subroutines, or code sections).

[0313] As an example, computer-executable instructions can be deployed to execute on a single electronic device, or on multiple electronic devices located at one location, or on multiple electronic devices distributed across multiple locations and interconnected via a communication network.

[0314] In summary, this application's embodiments analyze the historical transaction data set of the target resource account, calculate necessity scores using attribute features, and achieve accurate mapping of transaction data units to hierarchical categories such as survival categories. By quantifying the proportion of survival-related transactions, the rigid expenditure structure of the account is objectively reflected, and the determined resource stress level expands solvency assessment from a single resource dimension to a consumption structure dimension, significantly improving the accuracy of risk identification. The combined resource repayment strategy, which includes segmented repayment instructions and is generated by combining comprehensive scores, can balance the contradiction between resource repayment and maintaining basic survival needs when available resources are limited, avoiding the overdue risk caused by traditional single repayment models. By generating a resource composition distribution view and stress status dashboard components, the algorithm's decision-making is visualized, lowering the understanding threshold. The mechanism of batch execution of minimum amount repayment instructions or installment repayment instructions in response to confirmation signals ensures the atomicity and timeliness of multiple repayment actions, maximizing the credit value and liquidity efficiency of the target resource account.

[0315] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Any modifications, equivalent substitutions, and improvements made within the spirit and scope of this application are included within the scope of protection of this application.

Claims

1. A method for processing resource return, characterized in that, The method includes: In response to a resource return request for a target resource account, the system obtains a set of historical transaction data and the amount of available resources for the target resource account, wherein the set of historical transaction data includes multiple transaction data units. Based on the attribute characteristics of each transaction data unit, a necessity score for each transaction data unit is determined, and the multiple transaction data units are mapped to multiple hierarchical categories based on the necessity score, wherein the hierarchical categories include at least a survival category. Based on the transaction data units belonging to the survival category, the proportion of survival transactions is determined, and the resource pressure level of the target resource account is determined based on the proportion of survival transactions. Based on the necessity score and transaction attributes of each transaction data unit, a comprehensive score for each transaction data unit is determined. If the available resources are less than the total resources of the historical transaction data set, a combined resource return strategy containing segmented return instructions is generated based on the resource pressure level and the comprehensive score of each transaction data unit.

2. The method according to claim 1, characterized in that, The attribute features include merchant category identifiers and transaction summary information; determining the necessity score of each transaction data unit based on its attribute features includes: For each transaction data unit in the historical transaction data set, the merchant category identifier is matched based on a preset identifier mapping table, and the preset score corresponding to the matched merchant category identifier is determined as the basic score of the transaction data unit. Target keywords are extracted from the transaction summary information, and the semantic score of the transaction data unit is determined based on the matching degree between the target keywords and each of the hierarchical categories. The necessity score of the transaction data unit is obtained by weighting and summing the basic score and the semantic score according to the preset first weight ratio.

3. The method according to claim 2, characterized in that, The determination of the semantic score of the transaction data unit based on the matching degree between the target keywords and each of the hierarchical categories includes: Obtain a preset hierarchical keyword library, wherein the hierarchical keyword library includes a set of reference keywords corresponding to each of the hierarchical categories; For each of the aforementioned hierarchical categories, the number of times the target keyword is hit in the reference keyword set is counted. The ratio of the number of hits to the total number of reference keywords in the reference keyword set is determined as the matching degree between the target keyword and the hierarchical category; The target hierarchical category with the highest matching degree is selected from the multiple hierarchical categories, and the preset score corresponding to the target hierarchical category is determined as the semantic score of the transaction data unit.

4. The method according to claim 1, characterized in that, The mapping of the multiple transaction data units to multiple hierarchical categories based on the necessity score includes: Obtain a preset set of category division thresholds, wherein the set of category division thresholds includes at least a first boundary threshold for defining the survival category; For each transaction data unit in the historical transaction data set, if the necessity score of the transaction data unit is greater than or equal to the first threshold, the transaction data unit is mapped to the survival category.

5. The method according to claim 4, characterized in that, The set of category division thresholds also includes a second boundary threshold and a third boundary threshold, wherein the first boundary threshold is greater than the second boundary threshold, and the second boundary threshold is greater than the third boundary threshold; The mapping of the plurality of transaction data units to preset hierarchical categories based on the necessity score includes: For each transaction data unit in the historical transaction data set, if the necessity score of the transaction data unit is less than the first threshold and greater than or equal to the second threshold, the transaction data unit is mapped to the security category. If the necessity score of the transaction data unit is less than the second threshold and greater than or equal to the third threshold, the transaction data unit is mapped to the developmental category. If the necessity score of the transaction data unit is less than the third threshold, the transaction data unit is mapped to the enjoyment category.

6. The method according to claim 1, characterized in that, Determining the proportion of survival-type transactions based on the transaction data units belonging to the survival category includes: The resource amounts of the transaction data units belonging to the survival type category are summed to obtain the first total resource amount corresponding to the survival type category; The resource quantity of each transaction data unit in the historical transaction data set is summed to obtain the second total resource quantity corresponding to the historical transaction data set; The ratio of the first total resource quantity to the second total resource quantity is determined as the survival-type transaction proportion.

7. The method according to claim 6, characterized in that, The determination of the resource stress level of the target resource account based on the proportion of survival-type transactions includes: Obtain the periodic resource inflow of the target resource account and the number of historical overdue payments within a preset statistical period; The ratio of the second total resource amount to the periodic resource inflow amount is determined as the resource load ratio of the target resource account; The ratio of the number of historical overdue payments to the preset statistical period is determined as the normalized value of the number of historical overdue payments corresponding to the target resource account; According to the preset second weighting ratio, the proportion of survival-type transactions, the resource load ratio, and the normalized value of the number of historical overdue transactions are weighted and calculated to obtain the resource stress index of the target resource account. Based on the resource stress index, the resource pressure level of the target resource account is determined.

8. The method according to claim 7, characterized in that, Determining the resource stress level of the target resource account based on the resource stress index includes: Obtain a preset first tension threshold and a preset second tension threshold; If the resource stress index is greater than the first stress threshold, the resource stress level of the target resource account is determined to be high stress level. If the resource stress index is less than or equal to the first stress threshold and the resource stress index is greater than or equal to the second stress threshold, the resource pressure level of the target resource account is determined to be medium pressure level. If the resource stress index is less than the second stress threshold, the resource stress level of the target resource account is determined to be low stress level.

9. The method according to claim 1, characterized in that, The determination of a comprehensive score for each transaction data unit based on its necessity score and transaction attributes includes: For each of the transaction data units, the necessity score of the transaction data unit is normalized to obtain the necessity weight of the transaction data unit. The urgency weight of the transaction data unit is determined based on the transaction generation time and settlement reference time of the transaction data unit. The ratio of the resource quantity of the transaction data unit to the total resource quantity of the historical transaction data set is determined as the resource quantity percentage of the transaction data unit. Based on the historical resource return behavior of the target resource account, the preference coefficient corresponding to the transaction data unit is determined; According to the preset third weighting ratio, the necessity weight, the urgency weight, the resource quantity ratio, and the preference coefficient are weighted and summed to obtain the comprehensive score of the transaction data unit.

10. The method according to claim 9, characterized in that, The determination of the urgency weight of the transaction data unit based on the transaction generation time and settlement reference time includes: Determine the time difference between the settlement reference time and the transaction generation time of the transaction data unit; The ratio of the time difference to the settlement cycle duration corresponding to the transaction data unit is determined as the urgency weight of the transaction data unit.

11. The method according to claim 9, characterized in that, The step of determining the preference coefficient corresponding to the transaction data unit based on the historical resource return behavior of the target resource account includes: The system checks whether the target resource account has valid historical resource return records and obtains the detection results. If the detection result indicates that the target resource account has the historical resource return record, the frequency at which transactions belonging to the hierarchical category of the transaction data unit are prioritized is counted and used as the preference coefficient corresponding to the transaction data unit. If the detection result indicates that the target resource account does not have the historical resource return record, the corresponding default preference value is matched from the preset cold start preference mapping table as the preference coefficient of the transaction data unit according to the hierarchical category to which the transaction data unit belongs.

12. The method according to claim 1, characterized in that, The process of generating a combined resource return strategy, including segmented return instructions, based on the resource pressure level and the comprehensive score of each transaction data unit, includes: Based on the comprehensive score of each transaction data unit, the multiple transaction data units in the historical transaction data set are sorted to obtain a transaction data unit sequence; Based on the resource pressure level and the transaction data unit sequence, a combined resource return strategy containing segmented return instructions is generated. The segmented repayment instruction includes at least one of the following: minimum repayment instruction, full repayment instruction, and installment repayment instruction.

13. The method according to claim 12, characterized in that, The step of generating a combined resource return strategy containing segmented return instructions based on the resource pressure level and the transaction data unit sequence includes: Determine the total resource amount and minimum resource amount to be returned corresponding to the historical transaction data set; When the available resources are greater than or equal to the total resources, a first type of combined resource return strategy is generated based on the resource pressure level and the transaction data unit sequence. When the available resources are less than the total resources but greater than or equal to the minimum resource return amount, a second type of combined resource return strategy is generated based on a multidimensional evaluation model. If the available resources are less than the minimum amount of resources to be returned, a combined resource return strategy including a resource overdue warning is generated.

14. The method according to claim 13, characterized in that, Based on the resource pressure level and the transaction data unit sequence, a first type of combined resource return strategy is generated, including: When the resource pressure level is high, a first type of combined resource repayment strategy is generated, which includes the minimum repayment instruction. Wherein, when there is a target transaction data unit in the transaction data unit sequence with a resource amount greater than a preset resource amount and a necessity score greater than a preset score threshold, the first type of combined resource repayment strategy includes a single installment repayment instruction for the target transaction data unit. When the resource pressure level is medium pressure level, a first type of combined resource repayment strategy is generated, which includes the installment repayment instruction; wherein, when the resource proportion of transaction data units belonging to the enjoyment category in the transaction data unit sequence is greater than a first proportion threshold, the installment repayment instruction corresponds to a first installment period; when the resource proportion is between a second proportion threshold and the first proportion threshold, the installment repayment instruction corresponds to a second installment period. When the resource pressure level is low, a first type of combined resource return strategy is generated, which includes the full return instruction.

15. The method according to claim 13, characterized in that, The hierarchical categories also include: security category, development category, and enjoyment category; the second type of combined resource return strategy generated based on the multidimensional evaluation model includes: If the historical overdue normalized value of the target resource account within a preset statistical period is greater than or equal to a preset risk threshold, a second type of combined resource repayment strategy containing the full repayment instruction is generated. When the normalized value of the historical overdue period is less than the preset risk threshold, the proportion of survival-type transactions, the proportion of safe transactions corresponding to the safe category, the proportion of development-type transactions corresponding to the development category, and the proportion of enjoyment-type transactions corresponding to the enjoyment category of the historical transaction data set are weighted and summed according to the preset fourth weight ratio to obtain the transaction rationality index of the target resource account. The resource difference between the available resource quantity and the minimum repaid resource quantity corresponding to the historical transaction data set is determined, and the ratio of the resource difference to the necessary resource expenditure of the target resource account in a preset period is determined as the liquidity buffer rate of the target resource account. Based on the transaction rationality index and the liquidity buffer ratio, the phased return score of the target resource account is determined, wherein the phased return score is positively correlated with the transaction rationality index and negatively correlated with the liquidity buffer ratio; Based on the transaction rationality index, the liquidity buffer ratio, and the installment income score, a second type of combined resource return strategy is generated using a preset decision matrix.

16. The method according to claim 15, characterized in that, The second type of combined resource repayment strategy, generated using a preset decision matrix based on the transaction rationality index, the liquidity buffer ratio, and the phased return score, includes: The total installment service fee is determined based on the total resource amount corresponding to the historical transaction data set, the preset unit period service fee rate, and the preset number of installment periods. The ratio of the total installment service fee to the total resource amount is determined as the installment fee ratio. When the installment benefit score is greater than the installment fee ratio, it indicates that the installment repayment has positive benefits. When the transaction rationality index indicates high rationality and the liquidity buffer ratio indicates high pressure, a second type of combined resource repayment strategy is generated, which includes long-term installment repayment instructions. When the transaction rationality index indicates high rationality, the liquidity buffer ratio indicates moderate pressure, and installment repayment has positive returns, a second type of combined resource repayment strategy is generated, which includes medium- and long-term installment repayment instructions. When the transaction rationality index indicates high rationality, the liquidity buffer ratio indicates low pressure, and the phased return is positive, a second type of combined resource return strategy is generated, which includes an excess return instruction, wherein the return amount indicated by the excess return instruction is higher than the minimum return amount of resources. When the transaction rationality index indicates moderate rationality and the liquidity buffer ratio indicates high pressure, a second type of combined resource repayment strategy is generated, which includes the minimum repayment instruction and the fundraising prompt. When the transaction rationality index indicates low rationality, a second type of combined resource repayment strategy is generated, which includes the minimum repayment instruction and consumption optimization prompts.

17. The method according to claim 1, characterized in that, After generating the combined resource return strategy containing segmented return instructions, the method further includes: Based on the resource quantity statistics under each of the aforementioned hierarchical categories, a resource composition distribution view is generated. Based on the resource pressure level, a pressure status dashboard component is generated; The resource return feedback information, carrying the resource composition distribution view and the pressure status dashboard component, is returned to the terminal that sent the resource return request, so that the resource composition distribution view and the pressure status dashboard component can be displayed on the terminal.

18. The method according to claim 1, characterized in that, The segmented repayment instruction includes at least one of the following: minimum repayment instruction, full repayment instruction, and installment repayment instruction; after generating the combined resource repayment strategy containing the segmented repayment instruction, the method further includes: Receive an acknowledgment signal for the combined resource return strategy; In response to the confirmation signal, the resource agency return interface is invoked to execute in batches the minimum amount return instruction, the full amount return instruction, or the installment return instruction contained in the segmented return instruction.

19. A resource return processing device, characterized in that, The device includes: The data acquisition module is used to respond to a resource return request for a target resource account by acquiring the historical transaction data set and available resource quantity of the target resource account, wherein the historical transaction data set includes multiple transaction data units; The hierarchical category determination module is used to determine the necessity score of each transaction data unit based on the attribute characteristics of each transaction data unit, and to map the multiple transaction data units to multiple hierarchical categories based on the necessity score, wherein the hierarchical categories include at least the survival category. The resource pressure level determination module is used to determine the proportion of survival-type transactions based on the transaction data units belonging to the survival-type category, and to determine the resource pressure level of the target resource account based on the proportion of survival-type transactions. The comprehensive score determination module is used to determine the comprehensive score of each transaction data unit based on the necessity score and transaction attributes of each transaction data unit. The strategy generation module is used to generate a combined resource return strategy containing segmented return instructions based on the resource pressure level and the comprehensive score of each of the transaction data units, when the available resource amount is less than the total resource amount of the historical transaction data set.

20. An electronic device, characterized in that, The electronic device includes: Memory is used to store executable instructions or computer programs. A processor, when executing computer-executable instructions or computer programs stored in the memory, implements the resource return processing method according to any one of claims 1 to 18.

21. A computer-readable storage medium storing computer-executable instructions or a computer program, characterized in that, When the computer-executable instructions or computer program are executed by a processor, the resource return processing method according to any one of claims 1 to 18 is implemented.

22. A computer program product comprising computer-executable instructions or a computer program, characterized in that, When the computer-executable instructions or computer program are executed by a processor, the resource return processing method according to any one of claims 1 to 18 is implemented.

Images