Risk early warning and collection method for installment payment of equipment order

Abstract

Embodiments of the present application relate to the technical field of risk control, and particularly relate to a risk early warning and collection method for installment payment equipment order. Embodiments of the present application obtain order information, customer information, sales personnel information, repayment interaction information and customer follow-up records associated with overdue installment orders, generate processing flags and customer portrait information according to the customer follow-up records, carry out risk assessment on the overdue installment orders that need to be assessed, obtain a comprehensive risk score and a risk level, generate a differentiated collection strategy according to the comprehensive risk score, the risk level and the customer portrait information, and realize accurate identification, classified management and early warning of the risk of overdue installment orders, thereby improving the accuracy, comprehensiveness and timeliness of the risk assessment results, enabling timely measures to be taken, reasonably utilizing collection resources, improving the efficiency of collection resource allocation, enhancing the pertinence and effectiveness of the collection strategy, and reducing ineffective collection and collection costs.

Description

Technical Field

[0001] This application relates to the field of risk control technology, and in particular to a risk warning and collection method for installment payment equipment orders. Background Technology

[0002] Installment payment services are widely used in consumer finance, retail and supply chain fields. In order to reduce the risk of bad debts, the existing installment payment order management system adopts static rules based on a fixed overdue days threshold for risk assessment. That is, different risk levels are divided according to the overdue days of the order, and collection management is carried out accordingly.

[0003] Traditional installment payment risk control models, based on fixed repayment dates and cash flow monitoring, are suitable for conventional consumer finance scenarios involving the transfer of ownership of goods. However, in the sale of distributed energy and mobile communication equipment to emerging markets, there exists a special installment model of "pay-as-you-go": users are not purchasing ownership of the equipment, but rather a combination of "equipment usage rights + energy services." Equipment manufacturers control equipment access by issuing time-limited activation codes, using "service interruption" instead of "cash flow disruption" as the risk trigger. Under this model, once the activation codes run out, leading to service outages, users' willingness to repay deteriorates rapidly, and the traditional collection buffer period is lacking, resulting in an extremely narrow window for risk management. Existing risk control models based on fixed calendar dates are difficult to directly migrate to this special scenario. Traditional risk control solutions that are forcibly migrated to this special scenario will rely on the single indicator of overdue days for risk assessment, ignoring other dimensions of information about installment orders, and only triggering risk assessment after overdue payments occur, resulting in low accuracy and timeliness of risk assessment. Summary of the Invention

[0004] In view of this, one objective of the embodiments of this application is to provide a risk warning and collection method for installment payment equipment orders, so as to improve the low accuracy and timeliness of installment order risk assessment in related technologies.

[0005] In a first aspect, embodiments of this application provide a risk warning and collection method for installment payment device orders, comprising: acquiring original order data, including overdue installment orders and related order information, customer information, sales personnel information, repayment interaction information, and customer follow-up records; inputting customer follow-up records into a preset large language model, extracting state intent information and key objective information using the large language model, and integrating the state intent information and key objective information to obtain customer profile information, wherein the state intent information is used to reflect the customer's repayment willingness and level of communication cooperation, and the key objective information is used to reflect the customer's repayment ability and processing demands; generating a processing flag for overdue installment orders based on the customer profile information, wherein the processing flag is used to indicate the processing method for overdue installment orders; responding to the processing flag being a risk assessment flag, performing a risk assessment on the overdue installment orders based on the original order data to obtain a comprehensive risk score and risk level for the overdue installment orders, wherein the risk assessment flag is used to indicate the assessment of overdue installment orders; and generating a collection strategy for overdue installment orders based on the comprehensive risk score, risk level, and customer profile information, so that the target personnel can perform collection operations according to the collection strategy.

[0006] The embodiments of this application have the following beneficial effects: Unlike related technologies, the embodiments of this application obtain order information, customer information, sales personnel information, repayment interaction information, and customer follow-up records associated with overdue installment orders. Based on the customer follow-up records, processing flags and customer profile information are generated. Risk assessments are conducted on overdue installment orders that need to be evaluated to obtain a comprehensive risk score and risk level. Based on the comprehensive risk score, risk level, and customer profile information, differentiated collection strategies are generated. This achieves accurate identification, classification management, and early warning of risks associated with overdue installment orders, improves the accuracy, comprehensiveness, and timeliness of risk assessment results, allows for timely measures to be taken, makes reasonable use of collection resources, improves the efficiency of collection resource allocation, enhances the pertinence and effectiveness of collection strategies, and reduces ineffective collection and collection costs. Attached Figure Description

[0007] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the accompanying drawings used in the description of the related technologies or embodiments will be briefly introduced below. Obviously, the drawings described below only show some embodiments of this application and should not be considered as limiting the scope of protection. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0008] Figure 1 This is a schematic diagram illustrating an application scenario of the risk warning and collection method for installment payment equipment orders provided in some embodiments of this application; Figure 2This is a flowchart illustrating a risk warning and collection method for installment payment equipment orders provided in some embodiments of this application. Detailed Implementation

[0009] To make the objectives and advantages of the embodiments of this application more readily understood, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. The detailed description of the embodiments of this application in the accompanying drawings is not intended to limit the scope of protection claimed by this application, but only represents selected embodiments of this application. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0010] It should be noted that, unless there is a conflict, the various technical features involved in the embodiments of this application described below can be combined with each other, and all are within the protection scope of this application. Furthermore, although functional modules are divided in the device or structural schematic diagram and a logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the device or the order in the flowchart. In addition, the terms "first," "second," "third," and other similar expressions used herein do not limit the data or execution order, but are only for illustrative purposes and to distinguish identical or similar items with substantially the same function and effect, and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features.

[0011] Unless otherwise defined, the technical and scientific terms used in this specification have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the scope of this application. It should be understood that the term "and / or" as used in this specification includes any and all combinations of one or more of the listed items.

[0012] In conventional financial installment scenarios, users purchase ownership of goods in installments. Risk management hinges on monitoring user liquidity. Therefore, existing technologies typically employ risk control models based on fixed calendar dates, using "failure to fulfill monetary repayment obligations on the designated repayment date" as the sole criterion for delinquency. However, in the scenario of selling distributed energy to specific emerging markets, a special "pay-as-you-go" installment model exists, fundamentally different from conventional financial installments. Firstly, regarding the nature of the underlying asset, users do not purchase ownership but rather a combination of "equipment usage rights + energy services." Manufacturers control usage rights by issuing time-limited "equipment activation codes." Users receive activation permissions for the corresponding number of days after making periodic payments. The device verifies the remaining validity of the activation code in real time to control power supply. Permanent ownership is only granted after the user pays the full amount. This model is essentially "compensation in kind / services" rather than... The repayment method involves several key aspects. First, regarding the risk trigger mechanism, the risk trigger point for regular installment payments is "cash flow disruption," while the risk trigger point in this scenario is "service flow interruption," meaning the activation code is exhausted and not renewed, leading to device shutdown. Since the device directly serves the user's basic survival needs, the willingness to repay deteriorates rapidly after shutdown, and the risk of bad debts increases exponentially. Second, regarding the risk handling time window, users can still retain the goods after a regular installment order is overdue, providing a relatively long buffer period for collection. However, in this scenario, once the activation code is exhausted and shutdown is triggered, users face immediate inconvenience, and conventional mild collection strategies often fail, while aggressive measures are prone to escalating customer complaints. Therefore, the existing traditional risk control model based on "fixed repayment date" and "overdue funds" cannot be directly transferred to this special scenario. Forcibly transferring the traditional risk control solution to this special scenario would rely on the single indicator of overdue days for risk assessment, ignoring other dimensions of installment order information, and only triggering risk assessment after overdue occurs, reducing the accuracy and timeliness of risk assessment.

[0013] Based on this, this application provides a risk warning and collection method for installment payment equipment orders. By acquiring order information, customer information, sales personnel information, repayment interaction information, and customer follow-up records associated with overdue installment orders, processing flags and customer profile information are generated based on the customer follow-up records. A risk assessment is then conducted on overdue installment orders requiring evaluation to obtain a comprehensive risk score and risk level. Thus, differentiated collection strategies are generated based on the comprehensive risk score, risk level, and customer profile information. This achieves accurate identification, classification management, and early warning of overdue installment order risks, improving the accuracy, comprehensiveness, and timeliness of risk assessment results. Timely measures are taken, collection resources are used rationally, collection resource allocation efficiency is improved, and the targeting and effectiveness of collection strategies are enhanced, reducing ineffective collection and collection costs.

[0014] Please see Figure 1 , Figure 1 The illustration shows a schematic diagram of an application scenario for a risk warning and collection method for installment payment equipment orders provided by some embodiments of this application.

[0015] like Figure 1 As shown, this application scenario includes an electronic device 100 and a server 200, with the electronic device 100 communicating with the server 200 via a network. Examples of networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0016] The electronic device 100 includes any suitable type of device such as mobile device, terminal device, communication device, computing device, vehicle-mounted device, user equipment (UE), handheld device, and cloud device. For example, the electronic device 100 can be a desktop computer, laptop computer, all-in-one computer, tablet computer, microcontroller, and single-chip microcomputer, etc.

[0017] Server 200 includes any suitable type of server such as rack server, blade server, tower server, cabinet-type server, and micro server. Server 200 is used to store overdue installment orders and related order information, customer information, sales personnel information, repayment interaction information, and customer follow-up records (i.e., original order data).

[0018] For example, this application embodiment performs risk warning and collection operations for overdue installment orders: First, it obtains the original order data, for example, by obtaining the original order data from server 200 via the network. The original order data includes the overdue installment order and its associated order information, customer information, sales personnel information, repayment interaction information, and customer follow-up records. In some embodiments, order information, customer information, sales personnel information, repayment interaction information, and customer follow-up records can all be included in the overdue installment order.

[0019] Then, customer follow-up records are input into a pre-defined large language model. The model extracts status intent information and key objective information, which are then integrated to obtain customer profile information. Based on this profile, processing flags for overdue installment orders are generated. These flags indicate the processing method for overdue installment orders.

[0020] Next, it is determined whether the processing flag is a risk assessment flag. When the processing flag is a risk assessment flag, the overdue installment orders are risk assessed based on the original order data to obtain the comprehensive risk score and risk level of the overdue installment orders.

[0021] Finally, based on the comprehensive risk score, risk level, and customer profile information, a collection strategy for overdue installment orders is generated, enabling target personnel to execute collection operations according to the strategy. In some embodiments, after determining that the processing flag is not a risk assessment flag, it is determined whether the processing flag is a bad debt write-off flag. When the processing flag is a bad debt write-off flag, the overdue installment order is output to the manual review platform, allowing risk control personnel to execute write-off operations on the overdue installment orders in the manual review platform.

[0022] The above methods enable accurate identification, classification management, and early warning of risks associated with overdue installment orders. This improves the accuracy, comprehensiveness, and timeliness of risk assessment, allowing for timely action, rational use of collection resources, improved efficiency in resource allocation, enhanced targeting and effectiveness of collection strategies, and reduced ineffective collection and collection costs. Furthermore, when overdue installment orders are considered bad debts, they can be directly written off without risk assessment, saving assessment resources and improving their utilization rate.

[0023] It should be understood that Figure 1 The application scenario shown is merely an illustrative representation of one instance in which electronic device 100 is used to assess and warn of the risks of overdue installment orders and generate collection strategies in some embodiments of this application. Electronic device 100 is a laptop computer, and it does not impose any limitations on the structure, type, or quantity of electronic devices in other embodiments.

[0024] As can be understood from the above, the subject of implementation of the risk warning and collection method for installment payment equipment orders provided in this application embodiment can be any suitable type of electronic device with certain computing and control capabilities, such as the aforementioned electronic device 100.

[0025] The following will describe in detail the risk warning and collection method for installment payment device orders provided in this application embodiment, with reference to the exemplary application and implementation of the electronic device provided in the embodiments of this application.

[0026] For details, please refer to Figure 2 As shown, the risk warning and collection method for installment payment equipment orders provided in this application includes steps S31 to S34 to achieve warning and collection of overdue installment orders.

[0027] Step S31: Obtain the original order data.

[0028] The original order data includes overdue installment orders and related order information, customer information, sales personnel information, repayment interaction information, and customer follow-up records. Overdue installment orders refer to installment orders for the use of equipment that have already become overdue.

[0029] For example, in this application embodiment, a risk warning request regarding overdue installment orders is received, and in response to the risk warning request, the overdue installment orders to be processed are identified, and the order numbers of the overdue installment orders are obtained.

[0030] For example, the system accesses the order management system and retrieves the order information, customer information, sales personnel information, repayment interaction information, and customer follow-up records corresponding to the overdue installment order based on the order number.

[0031] For example, order information includes order number, product information, installment amount, number of installments, overdue amount, amount already paid, remaining amount due, number of overdue days, agreed repayment date, and order status.

[0032] The customer information includes the customer's name, contact information, identification, age, occupation, customer number, residential area, historical consumption records, historical installment records, and historical overdue records.

[0033] Among them, sales personnel information refers to the information of sales personnel responsible for overdue installment orders, including sales personnel number, department, area of ​​responsibility, contact information, sales amount, number of historical follow-ups, and historical collection success rate.

[0034] For example, repayment interaction information includes historical repayment time, amount and time of each repayment, payment channel, payment result, overdue repayment amount, and collection notice sending records.

[0035] Customer follow-up records refer to the interaction records generated between sales personnel and customers regarding repayment matters during the repayment management process for overdue installment orders. Customer follow-up records include the follow-up method and a summary of the communication content. The follow-up method records the communication channels, such as phone calls, emails, text messages, in-person visits, and social media contact. The communication content summary is a brief record of the core content and results of the communication in unstructured text format. For example, a communication content summary might be: "The customer claims this month's salary is delayed and requests a one-week grace period," "The customer promises to repay by next Friday," "The customer is resistant, claiming there is a problem with the product / goods and refusing to repay," or "Multiple attempts to contact them have failed; they appear to have lost contact."

[0036] In other embodiments, customer follow-up records also include the time of each follow-up, the person who followed up, the customer's response record, the follow-up results, and the handling opinions.

[0037] After obtaining order information, customer information, sales personnel information, repayment interaction information, and customer follow-up records, a data integrity check is performed on these data. If any information has a missing field, the corresponding information is retrieved again based on the order number until all information passes the data integrity check.

[0038] For example, after the order information, customer information, sales personnel information, repayment interaction information, and customer follow-up records pass the integrity verification, the order information, customer information, sales personnel information, repayment interaction information, and customer follow-up records that have passed the integrity verification are integrated in a unified data format and associated with the overdue installment orders to obtain the original order data, which is used for subsequent risk assessment, generation of processing flags, and customer profile information.

[0039] Step S32A: Input the customer follow-up records into the preset large language model, use the large language model to extract state intent information and key objective information, and integrate the state intent information and key objective information to obtain customer profile information.

[0040] Step S32B: Generate a processing flag for overdue installment orders based on customer profile information.

[0041] In this embodiment, the processing flag indicates the processing method for overdue installment orders. The processing method includes risk assessment and bad debt write-off. Risk assessment refers to evaluating the risk of overdue installment orders, and bad debt write-off refers to manually writing off overdue installment orders. Customer profile information includes status intent information and key objective information.

[0042] For example, when a salesperson adds or updates customer follow-up records for overdue installment orders in the order management system, a risk warning for overdue installment orders and the generation of collection strategies are triggered. Thus, this embodiment inputs all customer follow-up records associated with overdue installment orders into a large language model. The large language model performs deep semantic analysis on the customer follow-up records to identify the customer's expressed willingness and attitude to repay, and extracts status intent information. This status intent information includes at least one of the following: intention to commit to repayment, intention to become unreachable, intention to postpone repayment, intention to refuse repayment, intention to negotiate repayment, intention to continue communication, and intention to resolve disputes. Information is extracted from the factual descriptions in the customer follow-up records to identify the customer's objective situation (such as processing requests), obtaining key objective information. This key objective information includes at least one of the following: customer income, employment, cash flow, changes in contact information, family situation, historical default history, historical performance history, and medical conditions. The status intent information and key objective information are then linked and integrated to generate structured customer profile information. Status and intent information is used to reflect the customer's willingness to repay and the degree of cooperation in communication, while key objective information is used to reflect the customer's repayment ability and the handling of their requests.

[0043] For example, this embodiment categorizes status intent information into preset tags, such as "Normal Communication - Promised Repayment," "Financial Difficulty - Request for Extension," "Strong Resistance - Refusal to Communicate," "Out of Contact," and "Major Change (e.g., Death)." Key objective information refers to factual information about overdue installment orders, such as "the customer's promised next repayment date," "the specific reasons for the difficulties mentioned," and "objections to the product / goods."

[0044] For example, the system determines whether the status intent information in the customer profile includes an intention to commit to repayment (e.g., normal communication - commitment to repayment), an intention to negotiate repayment (e.g., financial hardship - request for extension), or an intention to refuse repayment (e.g., strong resistance - refusal to communicate). When the status intent information includes an intention to commit to repayment, an intention to negotiate repayment, or an intention to refuse repayment, a risk assessment flag is generated as a processing flag to instruct the overdue installment order to undergo risk assessment. When the status intent information includes an intention to be out of contact (e.g., out of contact, major change), a bad debt write-off flag is generated as a processing flag to instruct the overdue installment order to be manually written off.

[0045] For example, determine whether key objective information indicates that the customer is experiencing at least one of the following situations: income interruption, long-term unemployment, severe financial difficulties, invalid contact information, legal risk, or repeated defaults. If the key objective information indicates any of these situations, generate a risk assessment flag as the processing flag. When the status intent information and key objective information correspond to different processing results, the processing result of the status intent information shall prevail.

[0046] For example, the generated processing flag is output and attached to the overdue installment order so that the overdue installment order can be risk-assessed in response to the processing flag being a risk assessment flag, or the overdue installment order can be manually written off in response to the processing flag being a bad debt write-off flag.

[0047] In this embodiment, a processing flag is generated based on the processing status and risk changes in customer follow-up records. This flag determines whether overdue installment orders will proceed to risk assessment, continued follow-up, or bad debt write-off. Customer profile information is a set of customer characteristics formed based on customer follow-up behavior and communication feedback. It reflects the customer's repayment ability, willingness to repay, and level of cooperation, providing reliable data support for the accurate formulation of subsequent collection strategies.

[0048] Step S33: In response to the processing flag being a risk assessment flag, conduct a risk assessment on the overdue installment orders based on the original order data to obtain the comprehensive risk score and risk level of the overdue installment orders.

[0049] In this embodiment, the risk assessment flag is used to indicate the assessment of overdue installment orders.

[0050] For example, in response to overdue installment orders, the processing flag is a risk assessment flag, and risk characteristic data for risk assessment is extracted from the original order data. Risk characteristic data includes order information, customer information, sales personnel information, and repayment interaction information.

[0051] For example, order risk characteristics are extracted based on order information. These characteristics include total order amount, number of overdue days, remaining amount due, overdue amount, number of installments, repayment ratio, number of historical extensions, and order status.

[0052] Customer risk characteristics are extracted based on customer information. These characteristics include historical overdue payment frequency, historical default frequency, historical spending amount, historical installment success rate, customer's occupational stability, residential stability, and historical credit history.

[0053] Business risk characteristics are extracted based on sales personnel information. These characteristics include the sales personnel's historical collection success rate, the proportion of overdue orders handled, the default rate of the customers handled, and the proportion of historically risky orders handled by the sales personnel.

[0054] Risk characteristics of repayment behavior are extracted based on repayment interaction information. These characteristics include the interval between the most recent repayment date and the current date, the amount of overdue repayment, the historical on-time repayment rate, the number of consecutive non-payments, the fulfillment rate of promised repayments, the call connection rate, and the SMS response rate.

[0055] The risk characteristics of orders, customers, businesses, and repayment behavior are standardized to eliminate differences in dimensionality between different risk characteristics. The risk contribution value corresponding to each risk characteristic is calculated based on preset risk characteristic weights. Different risk characteristics correspond to different risk characteristic weights.

[0056] Finally, the risk contribution values ​​of each item are weighted and combined to obtain the comprehensive risk score for overdue installment orders. The higher the risk contribution value, the greater the impact of the corresponding risk characteristic on the order default risk.

[0057] For example, the overall risk score is matched with a preset risk level range. For instance, if the overall risk score is in the first risk range, the risk level of the overdue installment order is determined to be low; if the overall risk score is in the second risk range, the risk level of the overdue installment order is determined to be medium; if the overall risk score is in the third risk range, the risk level of the overdue installment order is determined to be high; and if the overall risk score is in the fourth risk range, the risk level of the overdue installment order is determined to be extremely high.

[0058] In this embodiment, the risk assessment of overdue installment orders is not based solely on a static judgment of the number of overdue days. Instead, it is a weighted assessment of multi-source risk characteristics from multiple dimensions, including order, customer, sales personnel, and repayment interaction. This approach can more accurately reflect the actual default risk of overdue installment orders, improve the accuracy of the overall risk score and risk level, and provide a reliable basis for developing differentiated collection strategies.

[0059] Step S34: Based on the comprehensive risk score, risk level, and customer profile information, generate a collection strategy for overdue installment orders so that the target personnel can carry out collection operations according to the collection strategy.

[0060] In this embodiment, customer profile information includes tags such as willingness to perform, ability to perform, communication and cooperation, risk of being out of contact, level of integrity, sensitivity to collection, and financial pressure.

[0061] The basic collection template is determined based on the risk level. Different risk levels correspond to different collection intensities; the higher the risk level, the higher the collection frequency, the stricter the collection measures, and the faster the collection escalation.

[0062] Based on customer profile information, the basic collection template was adjusted as follows: In response to the "willingness to repay" tag, indicating a high willingness to repay, reminder-style collection measures were prioritized; in response to the "financial pressure" tag, indicating financial difficulties, measures such as negotiated repayment, deferred repayment, or phased repayment were implemented; in response to the "communication and cooperation" tag, indicating a proactive customer, the proportion of telephone communication and manual follow-up was increased; in response to the "risk of losing contact" tag, indicating a risk of losing contact, the proportion of multi-channel outreach methods such as SMS, instant messaging, and emergency contact notifications was increased; and in response to the "low creditworthiness" tag, the collection frequency was increased, and risk escalation measures were implemented.

[0063] For example, collection parameters are adjusted based on the overall risk score of overdue installment orders. These parameters include collection frequency, collection interval, collection priority, collection channels, and collection escalation thresholds. The higher the overall risk score, the higher the collection frequency, the shorter the collection interval, and the higher the collection priority.

[0064] For example, target personnel are determined based on risk level and customer profile information. Target personnel include one or more of the following: sales personnel, customer service personnel, dedicated debt collectors, and legal personnel. When the risk level is low, sales personnel or customer service personnel are identified as target personnel; when the risk level is high, dedicated debt collectors are identified as target personnel; when the risk level reaches extremely high, dedicated debt collectors and legal personnel are identified as target personnel.

[0065] A collection plan is generated based on the collection parameters. The collection plan includes the initial collection date, subsequent collection timelines, collection channels for each collection, target personnel, and collection script types. Among them, collection script types include reminder scripts, negotiation scripts, warning scripts, and legal notification scripts.

[0066] For example, a collection strategy for overdue installment orders is generated by combining a basic collection template, adjusted collection measures, collection parameters, target personnel, and collection plan. The collection strategy is then pushed to the target personnel's corresponding terminal devices (e.g., mobile phones or computers), and collection tasks are generated according to the collection plan, so that the target personnel can perform at least one collection action according to the collection strategy, including telephone collection, SMS collection, instant message collection, in-person visit, or legal procedure notification.

[0067] In some embodiments, the comprehensive risk score, risk level, customer profile information, and prompts for generating collection strategies can be input into a large language model. Upon receiving the input data, the large language model performs deep semantic analysis on the comprehensive risk score, risk level, and customer profile information based on the prompts, generating and outputting personalized, highly targeted collection strategies or collection reports. Collection strategies include specific collection action suggestions and methods, while collection reports include collection strategies, customer status information, historical communication information, and risk assessment information.

[0068] After the target personnel complete the collection process, obtain the collection feedback results. The feedback results include whether the customer answered, the content of the customer's feedback, whether they promised to repay, the promised repayment date, the actual repayment status, and the outcome of this collection attempt. Update the customer follow-up record based on the collection feedback results to provide a basis for the next round of risk assessment and collection strategy generation.

[0069] In this embodiment, the collection strategy is not based solely on risk level. Instead, it uses a comprehensive risk score to determine the collection intensity and priority, risk level to determine the collection level and escalation path, and customer profile information to determine the collection method, collection channel, and collection script. This achieves differentiated and personalized configuration of collection strategies, improves the collection success rate, and reduces collection costs.

[0070] This application embodiment acquires order information, customer information, sales personnel information, repayment interaction information, and customer follow-up records associated with overdue installment orders. Based on the customer follow-up records, it generates processing flags and customer profile information. For overdue installment orders that require evaluation, it conducts risk assessments to obtain comprehensive risk scores and risk levels. Based on the comprehensive risk scores, risk levels, and customer profile information, it generates differentiated collection strategies to achieve accurate identification, classification management, and early warning of overdue installment order risks. This improves the accuracy, comprehensiveness, and timeliness of risk assessment results, enables timely measures to be taken, rationally utilizes collection resources, improves the efficiency of collection resource allocation, enhances the pertinence and effectiveness of collection strategies, and reduces ineffective collection and collection costs.

[0071] In some implementations, the embodiments of this application, through steps S33A to S33J, perform risk assessment on overdue installment orders based on the original order data to obtain the comprehensive risk score and risk level of the overdue installment orders.

[0072] Step S33A: Determine the score of the first order based on the total order amount, overdue amount, and overdue days.

[0073] The order information includes the total order amount, overdue amount, overdue days, installment period, and product value indicator. Customer information includes transaction activity index and historical overdue frequency. Sales personnel information includes sales bad debt rate. Repayment interaction information includes collection response rate and overdue repayment amount.

[0074] For example, based on the total order amount, overdue amount, and overdue days in the order information, the severity of overdue installment orders is assessed from two dimensions: the overdue amount ratio and the overdue time coefficient. The severity of overdue orders is then mapped to a standard score range. , and obtain the score for the first order.

[0075] In some embodiments, the present application implements steps S33A1 to S33A4 to determine the first order score based on the total order amount, the overdue amount, and the number of overdue days.

[0076] Step S33A1: Determine the overdue amount as the overdue ratio by the ratio of the overdue amount to the total order amount.

[0077] Step S33A2: Determine the time overdue coefficient based on the sum of the overdue days and the first preset value.

[0078] Step S33A3: The product of the overdue amount ratio and the overdue time coefficient is determined as the overdue severity coefficient.

[0079] For example, embodiments of this application are based on the following formula: Calculate the severity coefficient of overdue payments, where, The severity of the overdue period is the coefficient. For overdue amounts, Total order amount The percentage of overdue amounts. The number of overdue days The first preset value, , This is the time overdue coefficient.

[0080] For example, the overdue amount ratio is used to measure the severity of overdue installment orders in terms of "volume". For instance, an overdue amount of 100 yuan in an order with a total amount of 10,000 yuan (1% ratio) is significantly more serious than an order with a total amount of 1,000 yuan (10% ratio).

[0081] The time-delay coefficient measures the severity of overdue installment orders in terms of "time," depicting the non-linear growth law of diminishing marginal returns. The main purpose of the time-delay coefficient is to achieve "initial sensitivity, later desensitization." In the first few days of delinquency, the risk rises sharply; for example, from day 1 to day 3, the customer's willingness and likelihood of repayment change significantly. As the number of overdue days increases, the incremental risk with each additional day gradually decreases; for example, from day 100 to day 103, the change in risk status is no longer significant.

[0082] Understandably, this embodiment uses a logarithmic function (Log2) with base 2 to calculate the time lapse coefficient, which can provide good risk differentiation in risk assessment scenarios. In other embodiments, different bases can be selected according to actual needs, such as natural numbers. Or a logarithmic function with the base 10, etc., the core is to use the logarithmic function to simulate the diminishing marginal effect of risk growth.

[0083] Step S33A4: Obtain the first order score based on the overdue severity coefficient and the preset first overdue coefficient, second overdue coefficient, third overdue coefficient, first increase coefficient, second increase coefficient, and third increase coefficient.

[0084] Among them, engineers can customize the first overdue coefficient, second overdue coefficient, third overdue coefficient, first increase coefficient, second increase coefficient, and third increase coefficient based on experience data and actual needs.

[0085] This application's embodiment employs a piecewise function mapping method to calculate the continuous delinquency severity coefficients. Mapped to the standard score range The risk score is used to obtain the first order score. The core idea of ​​the piecewise function mapping method is that the risk score increases with different sensitivities for overdue installment orders of varying severity, in order to more accurately match the perception of business risk.

[0086] The piecewise function mapping method works by setting multiple sets of increasing coefficient threshold intervals and defining a corresponding linear transformation slope for each interval (i.e., pre-set first, second, and third amplification coefficients). The overdue severity coefficient is also considered. When the risk score is low, the increase per unit of growth is small (risk changes are gradual); however, when the delinquency severity coefficient is low... When entering the middle or higher threshold range, the increase in risk score per unit increase becomes greater (sensitive to risk changes); while when the delinquency severity coefficient... When the risk score is at its maximum, it approaches the maximum score of 100 (risk saturation).

[0087] For example, the coefficient threshold range and mapping rules for mapping the overdue severity coefficient using the piecewise function mapping method are shown in Table 1 below. Table 1 only illustrates the coefficient threshold range and mapping rules, and does not impose any limitations on the coefficient threshold range and mapping rules in other embodiments.

[0088] Table 1:

[0089] in, , , These are the first overdue coefficient, the second overdue coefficient, and the third overdue coefficient, as set in the embodiments of this application. , , .in , , These are the first amplification coefficient, the second amplification coefficient, and the third amplification coefficient, as set in the embodiments of this application. , , . This is the score for the first order. Among them, The first baseline score, ; As the second baseline score, .

[0090] According to Table 1, when the overdue severity coefficient... (i.e., located in the coefficient threshold range) When ), the score of the first order is calculated. When the severity coefficient of delinquency is... (i.e., located in the coefficient threshold range) When calculating the score of the first order, the score is determined. When the severity coefficient of delinquency is... (i.e., located in the coefficient threshold range) When calculating the score of the first order, the score is determined. When the severity coefficient of delinquency is... (i.e., located in the coefficient threshold range) When ), the score of the first order is calculated. .

[0091] In some embodiments, step S33A4 includes step S33A41.

[0092] Step S33A41: In response to the overdue severity coefficient being greater than or equal to zero and less than the first overdue coefficient, the product of the first increase coefficient and the overdue severity coefficient is determined as the first order score.

[0093] For example, according to Table 1, when the delinquency severity coefficient Greater than or equal to zero and less than the first delinquency coefficient When, that is, the severity coefficient of overdue payment. Located in the coefficient threshold range In this embodiment of the application, the first amplification coefficient will be... With overdue severity coefficient Multiply to get the score for the first order. ,Right now: .

[0094] In some embodiments, step S33A4 includes steps S33A42 to S33A44.

[0095] Step S33A42: In response to the fact that the delinquency severity coefficient is greater than or equal to the first delinquency coefficient and less than the second delinquency coefficient, determine the product of the first increase coefficient and the first delinquency coefficient as the first benchmark score.

[0096] Step S33A43: Multiply the difference between the second amplification coefficient and the first coefficient to obtain the product value of the first coefficient.

[0097] Step S33A44: The sum of the product of the first baseline score and the first coefficient is determined as the first order score.

[0098] The first coefficient difference is the difference between the overdue severity coefficient and the first overdue coefficient.

[0099] For example, according to Table 1, when the delinquency severity coefficient Greater than or equal to the first delinquency coefficient And less than the second delinquency coefficient At that time, i.e., the severity coefficient of overdue payment. Located in the coefficient threshold range In this application, the embodiments are based on the following formula: The score of the first order is calculated. , The first baseline score, . , These are the first and second growth coefficients, respectively. This is the difference in the first coefficient. This is the product value of the first coefficient.

[0100] In some embodiments, step S33A4 includes steps S33A45 to S33A47.

[0101] Step S33A45: In response to the fact that the delinquency severity coefficient is greater than or equal to the second delinquency coefficient and less than the third delinquency coefficient, determine the sum of the product of the first baseline score and the second coefficient as the second baseline score.

[0102] Step S33A46: Multiply the difference between the third amplification coefficient and the second coefficient to obtain the product value of the third coefficient.

[0103] Step S33A47: The sum of the product of the second baseline score and the third coefficient is determined as the first order score.

[0104] In this embodiment, the second coefficient product is the product of the second increase coefficient and the difference between the overdue coefficients, and the difference between the overdue coefficients is the difference between the second overdue coefficient and the first overdue coefficient. Specifically, the second coefficient difference is the difference between the overdue severity coefficient and the second overdue coefficient.

[0105] For example, according to Table 1, when the delinquency severity coefficient Greater than or equal to the second overdue coefficient And less than the third overdue coefficient At that time, i.e., the severity coefficient of overdue payment. Located in the coefficient threshold range In this application, the embodiments are based on the following formula: The score of the first order is calculated. , As the second baseline score, . , These are the first overdue coefficient and the second overdue coefficient, respectively. This is the difference in overdue coefficients. This is the product of the second coefficient. This is the difference in the second coefficient. This is the third growth factor. This is the product of the third coefficient.

[0106] In some embodiments, step S33A4 includes step S33A48.

[0107] Step S33A48: In response to the overdue severity coefficient being greater than or equal to the third overdue coefficient, determine the preset full score as the first order score.

[0108] For example, according to Table 1, when the delinquency severity coefficient Greater than or equal to the third overdue coefficient At that time, i.e., the severity coefficient of overdue payment. Located in the coefficient threshold range In this embodiment of the application, a preset maximum score of 100 is determined as the score for the first order. ,Right now: .

[0109] Step S33B: Determine the second order score based on the installment period.

[0110] For example, based on the installment period information in the order details, the installment length of overdue installment orders is assessed from the perspective of repayment cycle, and the installment length is mapped to a standard score range. , thus obtaining the second order score.

[0111] The calculation rules for the number of installments and the risk score are as follows: Based on the preset mapping rules, the number of installments (IN) is mapped to a standard score range. The risk score is based on the core principle of the mapping rule, which reflects the business consensus that "the longer the repayment period, the higher the future uncertainty and the increased risk of default."

[0112] In some embodiments, the present application implements steps S33B1 to S33B2 to determine the second order score based on the number of installments.

[0113] Step S33B1: Construct an installment formula based on the preset base installment amount, reference installment amount, minimum installment number, and maximum installment number.

[0114] In this embodiment, the phasing formula includes the number of phasing periods.

[0115] For example, the phased formula constructed in this application embodiment is as follows: , For the second order score, As the minimum number of installments, this application embodiment sets the minimum number of installments to 12, that is: Installment orders can be divided into at least 12 installments. The base installment percentage refers to the minimum number of installments. The corresponding installment amount is set at 20 in this embodiment of the application, that is: . As the maximum number of installments, this application embodiment sets the maximum number of installments to 54, that is: Installment orders can be divided into a maximum of 54 installments. The reference installment amount refers to the percentage of installments relative to the maximum installment amount. The corresponding installment amount is set to 80 in this embodiment of the application, that is: Among them, the benchmark installment score Compared with reference installment scores The sum of these is equal to the maximum score of 100. This refers to the number of periods in the phase.

[0116] Step S33B2: Substitute the installment period number into the installment formula to calculate the second order score.

[0117] For example, the number of installments Substituting into the installment formula constructed above, the score for the second order is calculated. .

[0118] For example, embodiments of this application set a number of phases. When the number of installments At that time, calculate the score of the second order. When the number of installments At that time, the score of the second order was calculated. When the number of installments At that time, the score of the second order was calculated. When the number of installments At that time, the score of the second order was calculated. .

[0119] Step S33C: Determine the third order score based on the product value indicator.

[0120] For example, based on the product value indicator in the order information, the value of overdue installment orders is assessed from the perspective of product value, and the product value indicator is mapped to a standard score range. The risk score is used to obtain the third order score.

[0121] For example, in this application embodiment, a commodity value indicator is established in advance and located within a standard score range. The correspondence between risk scores is shown in Table 2 below.

[0122] Table 2:

[0123] In Table 2, the product value indicators Sign-High, Sign-Medium, and Sign-Low correspond to risk scores of 100, 50, and 0, respectively. Analyzing the product value indicators, when the product value indicator is Sign-High, the risk score of 100 corresponding to Sign-High is determined as the third order score. ,Right now: When the commodity value indicator is Sign-Medium, the risk score of 50 corresponding to the commodity value indicator Sign-Medium is determined as the third order score. ,Right now: When the commodity value indicator is Sign-Low, the risk score of 0 corresponding to the commodity value indicator Sign-Low is determined as the third order score. ,Right now: .

[0124] The product value indicator can be determined by the order system based on the value of the product / goods when a customer places an order and generates a corresponding installment order, and the product value indicator is attached to the installment order. It is understood that Table 2 only schematically illustrates the correspondence between the product value indicator and the risk score, and does not impose any limitations on the product value indicator, risk score, or their correspondence in other embodiments.

[0125] Step S33D: Determine the first customer score based on the transaction activity index.

[0126] For example, based on the transaction activity index in customer information, the activity level of customers in overdue installment orders is assessed from the dimension of transaction activity, and the transaction activity index is mapped to a standard score range. The risk score is used to obtain the first customer score.

[0127] For example, in this application embodiment, a transaction activity index is pre-established and located within a standard score range. The correspondence between risk scores is shown in Table 3 below.

[0128] Table 3:

[0129] In Table 3, the trading activity indices DealActive-High, DealActive-Medium, and DealActive-Low correspond to risk scores of 20, 60, and 100, respectively. It should be understood that Table 3 is merely an illustrative representation of the relationship between the trading activity indices and risk scores, and does not impose any limitations on the trading activity indices, risk scores, or their corresponding relationships in other embodiments.

[0130] For example, a transaction activity index is determined based on a customer's transaction activity within a preset time period. For instance, if a customer's number of transactions within a first transaction time threshold (e.g., 90 days) is greater than or equal to a transaction frequency threshold (e.g., 3 times), or if the customer's most recent transaction occurred within a second transaction time threshold (e.g., 30 days), the transaction activity index is determined to be high (i.e., DealActive-High). If a customer's number of transactions within a third transaction time threshold (e.g., 180 days) is less than the transaction frequency threshold, or if the customer has no transactions within a fourth transaction time threshold (e.g., 150 days), the transaction activity index is determined to be low (i.e., DealActive-Low). It should be understood that any other suitable method can be used to determine the transaction activity index, and this application embodiment does not impose any specific limitations on this.

[0131] For example, when the trading activity index is DealActive-High, the risk score of 20 corresponding to the DealActive-High trading activity index is determined as the first customer score. ,Right now: When the trading activity index is DealActive-Medium, the risk score of 60, corresponding to the DealActive-Medium trading activity index, will be determined as the first client score. ,Right now: When the trading activity index is DealActive-Low, the risk score of 100 corresponding to the DealActive-Low index is determined as the first customer score. ,Right now: .

[0132] Step S33E: Determine the second customer score based on historical delinquency frequency.

[0133] For example, in this application embodiment, the number of historical installment orders corresponding to a customer is counted based on the historical installment records in the customer information. The number of historical overdue payments corresponding to a customer is also counted based on the historical installment records in the customer information. The ratio of the number of historical overdue payments to the number of historical installment orders is determined as the historical overdue frequency. Based on the historical overdue frequency, the overdue situation of customers in overdue installment orders is evaluated from the dimension of historical overdue frequency, and the historical overdue frequency is mapped to a standard score range. , thus obtaining the second customer score.

[0134] For example, this application embodiment uses a frequency-range scoring method to map historical overdue frequencies to standard score ranges. The non-linear mapping of the frequency interval scoring method reflects the business understanding that "the risk of low-frequency delinquency increases slowly, while the risk of medium- and high-frequency delinquency increases sharply."

[0135] For example, the overdue frequency ranges and mapping rules for mapping historical overdue frequencies according to the frequency range scoring method are shown in Table 4 below. Table 4 is only an illustrative representation of the overdue frequency ranges and mapping rules, and does not impose any limitations on the overdue frequency ranges and mapping rules in other embodiments.

[0136] Table 4:

[0137] In Table 4, Based on historical overdue frequency, The score for the second customer. , , These are the first frequency fraction, the second frequency fraction, and the third frequency fraction, as set in this embodiment. , , . , These are the first frequency threshold and the second frequency threshold, respectively, as set in the embodiments of this application. , . , , These are the reference frequency difference, the target frequency difference, and the standard frequency threshold, respectively. , , . For reference frequency scores, .in, This is a minimum value function used to find the minimum value.

[0138] Among them, when the historical overdue frequency At that time, calculate the second customer score. When the historical overdue frequency Values ​​greater than zero and less than or equal to the first frequency threshold (i.e., located in the overdue frequency range) When ), the second customer score is calculated: When the historical overdue frequency Greater than the first frequency threshold And less than or equal to the second frequency threshold (that is, it falls within the overdue frequency range) When ), the second customer score is calculated: When the historical overdue frequency Greater than the second frequency threshold (Located in the overdue frequency range) When ), the second customer score is calculated: .

[0139] In some embodiments, step S33E includes step S33E1.

[0140] Step S33E1: In response to the historical installment order count being zero, determine the preset conservative overdue score as the second customer score.

[0141] In this embodiment, customer information also includes the number of historical installment orders. For example, when the number of historical installment orders is zero, it indicates that the customer has never placed any installment orders in the past, making it impossible to calculate the historical delinquency frequency. Therefore, a pre-set conservative delinquency score is directly determined as the second customer score. The conservative delinquency score is set to 60, i.e.: .

[0142] Understandably, the purpose of setting a conservative overdue score of 60 is to implement the principle of prudent risk assessment. When there is insufficient data to make an accurate statistical judgment, it is necessary to avoid giving an evaluation that is too optimistic (i.e., low score) or extremely pessimistic (i.e., high score). A moderately high risk score (i.e., 60) is prudent and reasonable.

[0143] In some embodiments, step S33E includes step S33E2.

[0144] Step S33E2: In response to the historical overdue frequency being zero, the preset zero score is determined as the second customer score.

[0145] For example, when the number of historical installment orders is greater than zero, the historical delinquency frequency can be calculated. When the historical delinquency frequency is zero (i.e., ... When, according to Table 4, the preset zero score is determined as the second customer score in this embodiment of the application, that is: .

[0146] In some embodiments, step S33E includes step S33E3.

[0147] Step S33E3: In response to the historical overdue frequency being greater than zero and less than or equal to the first frequency threshold, the product of the first frequency ratio and the preset first frequency score is determined as the second customer score.

[0148] The first frequency ratio is the ratio of the historical overdue frequency to the first frequency threshold.

[0149] For example, according to Table 4, when the historical overdue frequency Values ​​greater than zero and less than or equal to the first frequency threshold (i.e., located in the overdue frequency range) When ), the embodiments of this application are based on the following formula: Calculate the second customer score , in the formula, The first frequency threshold, This is the first frequency ratio. The first frequency score, .

[0150] In some embodiments, step S33E includes steps S33E4 to S33E5.

[0151] Step S33E4: In response to the historical overdue frequency being greater than the first frequency threshold and less than or equal to the second frequency threshold, the ratio of the first frequency difference to the reference frequency difference is determined as the second frequency ratio.

[0152] Step S33E5: The sum of the first frequency score and the first frequency product is determined as the second customer score.

[0153] In this embodiment, the first frequency difference is the difference between the historical overdue frequency and the first frequency threshold, and the reference frequency difference is the difference between the second frequency threshold and the first frequency threshold. The first frequency product is the product of the second frequency ratio and a preset second frequency fraction.

[0154] For example, according to Table 4, when the historical overdue frequency Greater than the first frequency threshold And less than or equal to the second frequency threshold (that is, it falls within the overdue frequency range) When ), the embodiments of this application are based on the following formula: The second customer score is calculated. . In the formula, For reference frequency difference, .in The first frequency score, , The second frequency fraction, .

[0155] In the above formula, This is the first frequency difference. This is the second frequency ratio. This is the product value of the first frequency.

[0156] In some embodiments, step S33E includes steps S33E6 to S33E8.

[0157] Step S33E6: In response to the historical overdue frequency being greater than the second frequency threshold, the ratio of the second frequency difference to the target frequency difference is determined as the third frequency ratio.

[0158] Step S33E7: Determine the minimum value between the second frequency product and the third frequency fraction as the candidate score.

[0159] Step S33E8: The sum of the reference frequency score and the candidate score is determined as the second customer score.

[0160] In this embodiment, the second frequency difference is the difference between the historical overdue frequency and the second frequency threshold, and the target frequency difference is the difference between the standard frequency threshold 1 and the second frequency threshold.

[0161] In this embodiment, the second frequency product is the product of the third frequency ratio and the third frequency score. The third frequency score is the difference between a preset full score and a reference frequency score. The reference frequency score is the sum of the first frequency score and the second frequency score.

[0162] For example, according to Table 4, when the historical overdue frequency Greater than the second frequency threshold (Located in the overdue frequency range) When ), the embodiments of this application are based on the following formula: Calculate the second customer score . In the formula, For reference frequency scores, .in, The target frequency difference, . For the full score, The third frequency fraction, .

[0163] In the above formula, This is the second frequency difference. This is the third frequency ratio. This is the product value of the second frequency.

[0164] Step S33F: Determine the sales score based on the sales bad debt rate.

[0165] In this embodiment, the sales bad debt rate refers to the ratio of the total overdue amount of customers maintained / related by the sales personnel to the total order amount of customers maintained / related by the sales personnel. For example, if the total order amount is 200,000 yuan and the total overdue amount is 4,000 yuan, then the sales bad debt rate is... .

[0166] For example, in this embodiment of the application, based on the salesperson's ID, the installment orders of customers maintained / associated by the salesperson and the overdue orders therein are found. The total overdue amount and total order amount of customers maintained / associated by the salesperson are calculated, thereby calculating the sales bad debt rate. Based on the sales bad debt rate, this embodiment of the application assesses the creditworthiness of salespersons in overdue installment orders from the perspective of the sales bad debt rate, mapping the sales bad debt rate to a standard score range. , and obtain sales scores.

[0167] It is understandable that the core logic of the salesperson's bad debt rate range scoring method is: the higher the sales bad debt rate, the higher the final risk of the installment orders associated with / maintained by the salesperson, and therefore the higher the salesperson's credit risk score (i.e., sales score).

[0168] As an example, the bad debt rate range and mapping rules for mapping sales bad debt rates according to the bad debt rate range scoring method are shown in Table 5 below. Table 5 only illustrates the bad debt rate range and mapping rules and does not impose any limitations on the bad debt rate range and mapping rules in other embodiments.

[0169] Table 5:

[0170] In Table 5, For sales bad debt rate, Sales score. , , These are the first bad debt rate score, the second bad debt rate score, and the third bad debt rate score, as set in this embodiment. , , . , These are the first bad debt rate threshold and the second bad debt rate threshold, as set in this embodiment of the application. , . , , These are the reference bad debt rate difference, the target bad debt rate difference, and the standard bad debt rate threshold, respectively. , , . For reference, the bad debt ratio score, . This is a minimum value function used to find the minimum value.

[0171] Among them, when the sales bad debt rate At that time, the sales score was calculated. When the sales bad debt rate Values ​​greater than zero and less than or equal to the first bad debt rate threshold (i.e., within the bad debt rate range) When ), calculate the sales score: When the sales bad debt rate Greater than the first bad debt rate threshold And less than or equal to the second bad debt rate threshold. (that is, within the bad debt rate range) When calculating sales scores: When the sales bad debt rate Greater than the second bad debt rate threshold (within the overdue bad debt rate range) When ), the sales score is calculated: .

[0172] In some embodiments, step S33F includes step S33F1.

[0173] Step S33F1: In response to the number of maintenance customers being less than a preset threshold, determine the preset conservative overdue score as the sales score.

[0174] In this embodiment, the sales personnel information also includes the number of customers maintained. Engineers can customize and set a preset quantity threshold based on experience data and actual needs; for example, the preset quantity threshold could be set to 3.

[0175] For example, when the number of customers maintained is less than a preset threshold, it is considered that the sample size is insufficient, the sales bad debt rate is not calculated, and the preset conservative overdue score is directly determined as the sales score. The conservative overdue score is set to 60, i.e.: .

[0176] Understandably, this design is based on the principle of statistical robustness. When the sample size is insufficient, it avoids extreme or unreasonable calculations of bad debt rates due to small sample bias. For example, if one customer is overdue and the entire order amount is overdue, the bad debt rate would be 100%; if zero customers are overdue, the bad debt rate would be 0%. When the sample size is insufficient to make an accurate statistical judgment, it is important to avoid giving overly optimistic scores (i.e., low scores) or extremely pessimistic scores (i.e., high scores). A moderately high risk score (i.e., 60) is prudent and reasonable.

[0177] In some embodiments, step S33F includes step S33F2.

[0178] Step S33F2: In response to the number of customers being greater than or equal to a preset quantity threshold and the sales bad debt rate being zero, the preset zero score is determined as the sales score.

[0179] For example, when the number of customers maintained is greater than or equal to a preset threshold, and the sales bad debt rate is... Equal to zero (i.e.) and When this occurs, according to Table 5, in this embodiment of the application, the preset zero score is determined as the sales score, that is: .in, To maintain customer numbers, This is a preset quantity threshold.

[0180] In some embodiments, step S33F includes step S33F3.

[0181] Step S33F3: In response to the fact that the number of customers being maintained is greater than or equal to a preset quantity threshold and the sales bad debt rate is greater than zero and less than or equal to a first bad debt rate threshold, the product of the first bad debt rate ratio and the preset first bad debt rate score is determined as the sales score.

[0182] In this embodiment, the first bad debt ratio is the ratio of the sales bad debt ratio to the first bad debt ratio threshold.

[0183] For example, according to Table 5, when the number of customers maintained is greater than or equal to a preset threshold and the sales bad debt rate is... A value greater than zero and less than or equal to the first bad debt rate threshold (Right now and When ), the embodiments of this application are based on the following formula: Calculate the sales score . In the formula, The first bad debt rate threshold, This is the first bad debt ratio. The first bad debt ratio score,

[0184] In some embodiments, step S33F includes steps S33F4 to S33F5.

[0185] Step S33F4: In response to the fact that the number of customers being maintained is greater than or equal to a preset quantity threshold and the sales bad debt rate is greater than a first bad debt rate threshold and less than or equal to a second bad debt rate threshold, the ratio of the first bad debt rate difference to the reference bad debt rate difference is determined as the second bad debt rate ratio.

[0186] Step S33F5: The sum of the first bad debt rate score and the product of the first bad debt rate is determined as the sales score.

[0187] The first bad debt rate difference is the difference between the sales bad debt rate and the first bad debt rate threshold, and the reference bad debt rate difference is the difference between the second bad debt rate threshold and the first bad debt rate threshold. In this embodiment, the first bad debt rate product is the product of the second bad debt rate ratio and a preset second bad debt rate score.

[0188] For example, according to Table 5, when the number of customers maintained is greater than or equal to a preset threshold and the sales bad debt rate is... Greater than the first bad debt rate threshold And less than or equal to the second bad debt rate threshold. (Right now and When ), the embodiments of this application are based on the formula: Calculate sales score In the formula, The first bad debt ratio score, . The first bad debt rate threshold, This is the difference between the first bad debt rate and the second bad debt rate. To reference the difference in bad debt rates, and .in, This is the second bad debt ratio. The second bad debt ratio score, of which, , This is the product of the first bad debt ratio.

[0189] In some embodiments, step S33F includes steps S33F6 to S33F8.

[0190] Step S33F6: In response to the number of customers being maintained being greater than or equal to a preset quantity threshold and the sales bad debt rate being greater than a second bad debt rate threshold, the ratio of the second bad debt rate difference to the target bad debt rate difference is determined as the third bad debt rate ratio.

[0191] Step S33F7: Determine the minimum value between the second bad debt rate product and the third bad debt rate score as the undetermined score.

[0192] Step S33F8: The sum of the reference bad debt rate score and the pending score is determined as the sales score.

[0193] The second bad debt rate difference is the difference between the sales bad debt rate and the second bad debt rate threshold, and the target bad debt rate difference is the difference between the standard bad debt rate threshold 1 and the second bad debt rate threshold.

[0194] In this embodiment, the second bad debt rate product is the product of the third bad debt rate ratio and the third bad debt rate score. The third bad debt rate score is the difference between a preset full score and a reference bad debt rate score. The reference bad debt rate score is the sum of the first bad debt rate score and the second bad debt rate score.

[0195] For example, according to Table 5, when the number of customers maintained is greater than or equal to a preset threshold and the sales bad debt rate is... Greater than the second bad debt rate threshold (Right now and When ), the embodiments of this application are based on the formula: Calculate the sales score In the formula, For reference, the bad debt ratio score, . This is the second bad debt rate threshold. This is the difference in the second bad debt rate. The difference between the target bad debt rate and the target bad debt rate. .in, This is the third bad debt ratio. For the full score, The third bad debt ratio score, of which, , This is the product of the second bad debt rate.

[0196] Step S33G: Determine the score for the first line based on the collection response rate.

[0197] For example, based on the collection response rate in the repayment interaction information, the response of customers in overdue installment orders is evaluated from the dimension of collection response, and the collection response rate is mapped to a standard score range. The risk score is used to obtain the score for the first line.

[0198] For example, in this application embodiment, the collection response rate and its position within the standard score range are established in advance. The correspondence between risk scores is shown in Table 6 below.

[0199] Table 6:

[0200] As shown in Table 6, the DebtResponse-ExtremelyHigh, DebtResponse-High, DebtResponse-Medium, and DebtResponse-Low response levels correspond to risk scores of 0, 40, 70, and 100, respectively. It is understood that Table 6 is merely an illustrative representation of the correspondence between the response level and the risk score, and it does not impose any limitations on the response level, risk score, or their correspondence in other embodiments.

[0201] For example, the collection response level is determined based on the customer's actual feedback (e.g., repayment, partial repayment, promised repayment but not repaid, no response) within a preset number of days (e.g., 7 days) after the most recent collection request. For instance, if the customer has repaid within the preset number of days after the collection request, the collection response level is determined to be extremely high (i.e., DebtResponse-ExtremelyHigh). If the customer partially repays within the preset number of days after the collection request, the collection response level is determined to be high (i.e., DebtResponse-High). If the customer promises repayment but fails to repay within the preset number of days after the collection request, the collection response level is determined to be medium (i.e., DebtResponse-Medium). If the customer does not respond within the preset number of days after the collection request, the collection response level is determined to be low (i.e., DebtResponse-Low). It should be understood that any other suitable method can be used to determine the collection response level, and this application embodiment does not limit this in any way.

[0202] For example, when the collection response score is DebtResponse-ExtremelyHigh, the risk score of 0 corresponding to the collection response score DebtResponse-ExtremelyHigh is determined as the first row score. ,Right now: When the collection response rate is DebtResponse-High, the risk score of 40 corresponding to DebtResponse-High is determined as the first row score. ,Right now: When the debt collection response level is DebtResponse-Medium, the risk score of 60 corresponding to DebtResponse-Medium will be determined as the first-line score. ,Right now: When the collection response rate is DebtResponse-Low, the risk score of 100 corresponding to the collection response rate DebtResponse-Low is determined as the first row score. ,Right now: .

[0203] Understandably, when a customer promises to repay but fails to do so within the promised period, the collection response score will be updated from DebtResponse-Medium to DebtResponse-Low after the promised repayment period ends, so that the score in the first line is adjusted to the risk score of 100 corresponding to the collection response score DebtResponse-Low.

[0204] Step S33H: Determine the score for the second line based on the overdue payment amount.

[0205] The overdue payment amount refers to the actual amount repaid by the customer after overdue collection efforts. For example, based on the overdue payment amount in the repayment interaction information, the customer's repayment situation in overdue installment orders is assessed from the perspective of repayment amount, and the overdue payment amount is mapped to a standard score range. The risk score is used to obtain the second line score.

[0206] For example, in this embodiment of the application, the repayment ratio after overdue collection is calculated based on the overdue amount and the overdue amount, that is, the repayment ratio is the ratio of the overdue amount to the overdue amount. A repayment ratio range and a standard score range are established in advance. The correspondence between risk scores is shown in Table 7 below.

[0207] Table 7:

[0208] According to Table 7, the repayment ratio range , , , These correspond to risk scores of 100, 70, 40, and 0, respectively. It is understood that Table 7 is merely an illustrative representation of the correspondence between repayment ratio ranges and risk scores, and it does not impose any limitations on repayment ratio ranges, risk scores, or their correspondence in other embodiments.

[0209] When the repayment ratio is within the repayment ratio range At that time, determine the repayment ratio range The corresponding risk score of 100 is the score for the second row. ,Right now: When the repayment ratio is within the repayment ratio range At that time, the repayment ratio range will be determined. The corresponding risk score of 70 is the score for the second row. ,Right now: When the repayment ratio is within the repayment ratio range At that time, the repayment ratio range will be determined. The corresponding risk score of 40 is the score for the second row. ,Right now: When the repayment ratio is within the repayment ratio range At that time, determine the repayment ratio range The corresponding risk score of 0 is the score for the second row. ,Right now .

[0210] In some embodiments, this application embodiment uses steps S33H1 to S33H2 to determine the score of the second line based on the overdue repayment amount.

[0211] Step S33H1: In response to the ratio of overdue repayment amount to overdue amount being greater than or equal to a preset ratio threshold, a preset zero score is determined as the score for the second row.

[0212] Step S33H2: In response to the ratio of overdue repayment amount to overdue amount being less than the ratio threshold, determine the preset full score as the score for the second row.

[0213] In this embodiment, the overdue repayment amount refers to the actual repayment amount of the overdue installment order (i.e., the actual amount repaid by the customer after overdue collection). Engineers can customize the ratio threshold according to actual needs.

[0214] For example, when the ratio of the overdue payment amount to the overdue amount (i.e., the repayment ratio) is greater than or equal to a preset ratio threshold (such as 60%), a preset zero score is determined as the score for the second row. ,Right now: When the ratio of overdue payment amount to overdue amount (i.e., repayment ratio) is less than a preset ratio threshold, a preset maximum score of 100 is determined as the score for the second row. ,Right now: .

[0215] Step S33I: Weight and merge the scores of the first order, the second order, the third order, the first customer, the second customer, the sales, the first line, and the second line to obtain the comprehensive risk score.

[0216] For example, in this embodiment of the application, the preset first order score, second order score, third order score, first customer score, second customer score, sales score, first behavior score, and the fusion weight corresponding to the second behavior score are stored in local storage. The first order score, second order score, third order score, first customer score, second customer score, sales score, first behavior score, and the fusion weight corresponding to the second behavior score are obtained from the local storage. Each score is multiplied by the fusion weight corresponding to each score to obtain each weight score value. All weight score values ​​are added together to obtain the comprehensive risk score.

[0217] In some embodiments, step S33I includes steps S33I1 to S33I1A.

[0218] Step S33I1: Obtain multiple target weights.

[0219] In this embodiment, the multiple target weights include the first order weight corresponding to the first order score, the second order weight corresponding to the second order score, the third order weight corresponding to the third order score, the first customer weight corresponding to the first customer score, the second customer weight corresponding to the second customer score, the sales weight corresponding to the sales score, the first row weight corresponding to the first row score, and the second row weight corresponding to the second row score.

[0220] For example, in this application embodiment, completed historical orders are used as sample installment orders. The final business performance of these sample installment orders is used as the optimization objective. Through constrained mathematical optimization methods, optimal weight parameters (including first order weight, second order weight, third order weight, first customer weight, second customer weight, sales weight, first behavior weight, and second behavior weight) are trained and obtained, thereby enabling data-driven self-evolution capabilities. Here, completed historical orders refer to installment orders that were previously overdue but have since been completed.

[0221] In some embodiments, step S33I1 includes steps S33I11 to S33I17.

[0222] Step S33I11: Obtain target order data.

[0223] The target order data includes multiple sample installment orders that have been overdue and completed. Each sample installment order is associated with order information, customer information, sales personnel information, and repayment interaction information.

[0224] For example, in this application embodiment, the order management system is accessed, historical completed orders are obtained from the order management system as sample installment orders, and order information, customer information, sales personnel information and repayment interaction information associated with the sample installment orders are obtained.

[0225] Step S33I12: Based on multiple initial weights and the order information, customer information, sales personnel information and repayment interaction information associated with the candidate installment orders, conduct a risk assessment on the candidate installment orders to obtain a comprehensive risk score for the candidate installment orders.

[0226] In this step, the candidate installment order is any one of the multiple sample installment orders. Multiple initial weights include the initial weight corresponding to the first order score, the initial weight corresponding to the second order score, the initial weight corresponding to the third order score, the initial weight corresponding to the first customer score, the initial weight corresponding to the second customer score, the initial weight corresponding to the sales score, the initial weight corresponding to the first row score, and the initial weight corresponding to the second row score.

[0227] For example, using a method similar to steps S33A to S33H, a risk assessment is performed on the candidate installment order based on the associated order information, customer information, sales personnel information, and repayment interaction information, resulting in the following scores for each candidate installment order: first order score, second order score, third order score, first customer score, second customer score, sales score, first behavior score, and second behavior score. Each of these scores is multiplied by its corresponding initial weight: first order initial weight, second order initial weight, third order initial weight, first customer initial weight, second customer initial weight, sales initial weight, first behavior initial weight, and second behavior initial weight, respectively, to obtain a weighted score. All weighted scores are then summed to obtain the overall risk score for the candidate installment order.

[0228] Step S33I13: Based on the order information, customer information, sales personnel information, and repayment interaction information associated with the candidate installment orders, construct the result array corresponding to the candidate installment orders.

[0229] The result array includes a first value, a second value, and a third value. The first value represents the processing result of the candidate installment order, the second value represents the repayment period of the candidate installment order, and the third value represents the collection cost-benefit ratio of the candidate installment order. The repayment period refers to the number of calendar days from the first overdue date to the final settlement date. It should be understood that the collection cost-benefit ratio is the ratio of collection cost to the amount recovered. Collection costs include human resources costs, transportation costs, communication costs, etc., while the amount recovered refers to the final actual repayment amount of the candidate installment order.

[0230] For example, the result array is , where the first parameter To characterize candidate installment orders The parameters of the processing result, the first parameter The value (i.e., the first value) is used to specifically represent the candidate installment order. The processing results, for example Indicates candidate installment orders Write-off as bad debt Indicates candidate installment orders The debt has been successfully settled (including normal settlement and early settlement).

[0231] Second parameter To characterize candidate installment orders The parameter for payment collection time, the second parameter The value (i.e., the second value) is used to specifically represent the candidate installment order. The repayment period, for example Indicates candidate installment orders The repayment period is 21 days. For candidate installment orders that have been written off as bad debts, The value is assigned to a maximum constant representing "infinity". (For example ).

[0232] Third parameter To characterize candidate installment orders The third parameter is the cost-benefit ratio parameter for debt collection. The value (i.e., the third value) is used to specifically represent the candidate installment order. The cost-benefit ratio of debt collection is calculated, for example, if the cost of collection is 200 yuan and the amount recovered is 5000 yuan. , indicating candidate installment orders The cost-benefit ratio of collection is It's understandable that the lower the cost-benefit ratio of debt collection, the higher the efficiency of debt collection; conversely, the higher the cost-benefit ratio, the lower the efficiency of debt collection. The same applies to candidate installment orders that have been written off as bad debts. The value is assigned to a maximum constant representing "infinity". (For example ).

[0233] Step S33I14: Based on the result array and the comprehensive risk score of the candidate installment orders, calculate the result classification loss, payment time loss and cost-benefit ratio loss corresponding to the candidate installment orders.

[0234] Step S33I15: Weight the loss from the result classification, the loss from the collection period, and the loss from the cost-benefit ratio to obtain the comprehensive loss corresponding to the candidate installment order.

[0235] Step S33I16: Iteratively update multiple initial weights based on the comprehensive loss corresponding to all candidate installment orders until the comprehensive loss corresponding to all candidate installment orders is minimized.

[0236] Step S33I17: Determine the initial weights corresponding to the minimum comprehensive loss of all candidate installment orders as the final target weights.

[0237] It is understood that the goal of the weight update in this application is to find a new set of weights. The comprehensive risk score is calculated using this set of weights. It can simultaneously and best perform the following three prediction tasks: ①. Task L1 (classification): The size can clearly distinguish whether an order is written off as a bad debt (i.e. ) or determined to be settled (i.e. ); ②. Task L2 (Regression): Size and repayment time Positive correlation (i.e., the higher the overall risk score, the longer the expected recovery time); ③. Task L3 (Regression): Size and cost-effectiveness ratio of collection Positive correlation (i.e., the higher the overall risk score, the higher the cost-benefit ratio of collection, and the lower the collection benefit, the higher the collection cost).

[0238] in, These are the weights for the first order, the second order, the third order, the first customer, the second customer, the sales, the first behavior, and the second behavior.

[0239] This application's embodiments transform "updating weights" into a mathematical optimization problem. By constructing a multi-objective loss function that integrates classification and regression tasks and imposing business constraints, it is possible to automatically learn from the data the optimal comprehensive risk score. A weighted configuration that simultaneously considers "discrimination," "timeliness sensitivity," and "cost sensitivity" is employed. To accomplish these three prediction tasks concurrently, a comprehensive loss function is constructed. This guides the adjustment of weights and bias parameters. (Comprehensive Loss Function) It is a weighted sum of three sub-loss terms, each of which calculates the average prediction error of the current weight parameters over the entire training set.

[0240] In this embodiment, the comprehensive loss function is: .

[0241] in, This refers to the classification loss of the candidate installment order, used to measure the current weight. With bias parameters The degree of error in the overall prediction of "whether a debt will become bad." First, based on the comprehensive risk score. With bias parameters Calculate candidate installment orders The probability of being predicted as a bad debt ,in, It is a natural constant. Secondly, the classification loss function is used. Calculate the mean cross-entropy loss for all samples, where the calculation formula is: . For candidate installment orders The actual processing result label (bad debt or settlement), For the number of candidate installment orders, For predicted candidate installment orders The probability of bad debts. This refers to summation. It can be understood that the smaller the value of the classification loss function, the stronger its ability to distinguish between bad debt orders and normally settled orders.

[0242] Understandably, yes The function is used to divide the comprehensive risk into any range. (a real number) passes through Function mapping to The interval is interpreted as the probability value that a candidate installment order is predicted to be a "bad debt". In actual business operations, The larger, The closer it is to 1 (100% probability of bad debt); The smaller, The closer it is to 0, the better. (Bias parameter) It serves as an overall calibration tool.

[0243] Cross-entropy loss is a classic loss function in machine learning used for binary classification tasks (in this case, "whether it is a bad debt"). Its core function is to quantify the "gap" or "error level" between the predicted probability and the actual situation of the candidate installment order (i.e., the true processing result label).

[0244] In this embodiment of the application, candidate installment orders are first output. The probability of being classified as "bad debt" (A number between 0 and 1), while the actual processing result label It is a definite "yes" or "no" label (mathematically represented by 1 or 0). Cross-entropy is calculated using the formula... Calculate the loss for a single sample (i.e., candidate installment orders). If the candidate installment orders... Settle (i.e.) The classification loss function simplifies to If the candidate installment order The actual situation is bad debt (i.e. The classification loss function simplifies to The probability of prediction The higher the value (the closer it is to 1), the greater the loss value. The lower the value, the higher the loss value will be if the model makes a wrong prediction (predicting a very low probability). The purpose of designing the above classification loss function is to adaptively adjust the parameters (including weights) during the weight update and optimization process. and bias parameters This makes the predicted probability... The labels should be as close as possible to the actual processing results to achieve accurate classification.

[0245] Classification loss function The calculation process is as follows: First, for each candidate installment order in the sample training set... The candidate installment orders are calculated using the cross-entropy formula described above. Individual loss values. Then, all of them... The sum of losses for each candidate installment order is calculated to obtain the total loss. Finally, the total loss is divided by the number of candidate installment orders. This yields the average of the losses for all samples, known as the "mean". This mean is a single scalar number that represents the current weights. and bias parameters The average, overall classification error across the entire training set.

[0246] in, This refers to the regression loss based on the repayment period for candidate installment orders, used to measure the overall risk score. With repayment time The overall deviation. The formula for calculating the regression loss based on payment collection time is: .in, It is the squared error of a single sample (candidate installment order). First, calculate the comprehensive risk score. With repayment time The difference is then squared. Squaring ensures that the error is always positive and imposes a heavier penalty on larger errors. This is understandable. The smaller the value, the more accurate the prediction of the repayment time.

[0247] This refers to the cost-benefit ratio regression loss corresponding to candidate installment orders, used to measure the overall risk score. Cost-effectiveness ratio of collection The overall deviation. The formula for calculating the cost-benefit ratio regression loss is: .in It is the squared error of a single sample (candidate installment order). First, calculate the comprehensive risk score. Cost-effectiveness ratio of collection The difference is then squared. Squaring ensures that the error is always positive and imposes a heavier penalty on larger errors. This is understandable. The smaller the value, the more accurate the prediction of the cost-benefit ratio of debt collection.

[0248] in, , , These are all configurable hyperparameters used to balance the relative importance of the three business objectives: "reducing bad debts," "accelerating receivables collection," and "controlling costs." Engineers can adjust the hyperparameters according to actual needs. , , To flexibly adapt to the business strategy priorities of different periods.

[0249] For example, solving under constraints: the goal of weight update is to find a set of weights. With bias parameters Under the following nonnegativity and normalization constraints, the comprehensive loss function is such that... The value of (i.e., the comprehensive loss corresponding to all candidate installment orders) is minimized on the entire sample training set.

[0250] The nonnegativity constraint is as follows: All are greater than or equal to 0. The normalization constraint is: To maintain a total weight of 100%, the risk scores calculated under different weight versions are stable and comparable.

[0251] For example, this application embodiment uses the projected gradient descent optimization algorithm to iteratively update the weights using sample installment order data. With bias parameters Until the comprehensive loss function Satisfying the convergence condition (i.e., the comprehensive loss function) (minimum value) The weight combination obtained at convergence This refers to the updated optimal parameters, which determine the multiple initial weights (i.e., weight combinations) corresponding to minimizing the overall loss for all candidate installment orders. ) represents the final weights for multiple objectives.

[0252] In some embodiments, the weight combination is tested on a validation dataset independent of the sample training set. The performance of the dataset is evaluated. The validation dataset includes completed historical installment order data that is located after the training set in the time series, used to simulate the new weight combination. Performance on future order data. Specifically, when new weight combinations are used... If the weight combination does not significantly worsen business metrics (including classification loss and regression loss) and performs at the same level or better, it is considered a weight combination. If the overall performance is not lower than the current weight, the verification is passed and the device is entered into the deployment queue.

[0253] Next, the weights are combined using a phased, gradual rollout strategy (e.g., applying new weights to specific regions first). Update the overall risk score for overdue installment orders using updated weights.

[0254] In this embodiment, the weight combination is monitored in real time. Overall performance in real-world usage environments. Monitoring metrics include, but are not limited to: bad debt conversion rate for high-risk orders, differentiation between good and bad orders, and overdue payment trends for orders at each risk level. When core business metrics show a statistically significant deterioration compared to historical baseline levels, or trigger a preset absolute risk threshold, or receive a manual rollback instruction from the administrator, a rollback mechanism is triggered to restore the weight configuration to the previous stable version, ensuring the stability and reliability of online risk monitoring and early warning.

[0255] Step S33I2: The product of the first order weight and the first order score is determined as the first intermediate score.

[0256] Step S33I3: The product of the second order weight and the second order score is determined as the second intermediate score.

[0257] Step S33I4: The product of the third order weight and the third order score is determined as the third intermediate score.

[0258] Step S33I5: The product of the first customer weight and the first customer score is determined as the fourth intermediate score.

[0259] Step S33I6: The product of the second customer weight and the second customer score is determined as the fifth intermediate score.

[0260] Step S33I7: Determine the sixth intermediate score by multiplying the sales weight by the sales score.

[0261] Step S33I8: The product of the weight of the first row and the score of the first row is determined as the seventh intermediate score.

[0262] Step S33I9: The product of the weight in the second row and the score in the second row is determined as the eighth intermediate score.

[0263] Step S33IA: Sum the first, second, third, fourth, fifth, sixth, seventh, and eighth median scores to obtain the overall risk score.

[0264] For example, after obtaining multiple target weights, the comprehensive risk score of overdue installment orders is calculated according to steps S33I2 to S33IA.

[0265] Step S33J: Determine the risk level of overdue installment orders based on the comprehensive risk score.

[0266] The risk level of overdue installment orders is determined based on the comprehensive risk score, including steps S33J1 to S33J3.

[0267] Step S33J1: In response to the comprehensive risk score being greater than or equal to zero and less than a preset first score threshold, determine the low risk level as the risk level of the overdue installment order.

[0268] Step S33J2: In response to the comprehensive risk score being greater than or equal to the first score threshold and less than the preset second score threshold, determine the medium risk level as the risk level of the overdue installment order.

[0269] Step S33J3: In response to the overall risk score being greater than or equal to the second score threshold, the high-risk level is determined as the risk level of the overdue installment order.

[0270] It is understood that engineers can customize the first and second score thresholds based on experience data and actual needs, and this application embodiment does not impose any limitations on this. For example, the first score threshold... The second score threshold is .

[0271] When the comprehensive risk score Greater than or equal to zero and less than the first score threshold (Right now When determining the risk level of an overdue installment order, a low-risk level is assigned. When the comprehensive risk score... Greater than or equal to the first score threshold And less than the second score threshold (Right now If the risk level is 0, then the medium risk level is determined as the risk level of the overdue installment order. When the comprehensive risk score... Greater than the second score threshold (Right now If the risk level is high, then the risk level of the overdue installment order is determined.

[0272] In some embodiments, the risk warning and collection method for installment payment equipment orders provided in this application further includes step S35.

[0273] Step S35: In response to the processing flag being the bad debt write-off flag, output the overdue installment orders to the manual review platform, so that the risk control personnel can perform write-off operations on the overdue installment orders in the manual review platform.

[0274] In this embodiment, the bad debt write-off flag is used to indicate the write-off of overdue installment orders. For example, when the processing flag of an overdue installment order is the bad debt write-off flag, the overdue installment order is output to the manual review platform, and the risk control personnel are notified so that the risk control personnel can read the overdue installment order from the manual review platform and perform the write-off operation on the overdue installment order.

[0275] In some implementations, the embodiments of this application, through steps S341 to S344, generate a collection strategy for overdue installment orders based on comprehensive risk scores, risk levels, and customer profile information, so that the target personnel can perform collection operations according to the collection strategy.

[0276] Step S341: Input the comprehensive risk score, risk level and customer profile information into the preset big language model so that the big language model can generate collection strategies or collection reports.

[0277] The target personnel include one or more of the sales staff and sales managers.

[0278] Specifically, when the risk level is low or medium, the big language model generates a collection strategy; when the risk level is high, the big language model generates a collection report, which includes the collection strategy, customer status information, historical communication information, and risk assessment information.

[0279] For example, the model input data is constructed by organizing the comprehensive risk score, risk level, and customer profile information. The model input data includes at least the risk assessment results (i.e., the comprehensive risk score and risk level), customer profile information, and basic order information.

[0280] The process involves encapsulating model input data according to a preset prompt template to generate large language model input instructions. The preset prompt template defines the output content and format of the large language model. These instructions are then input into the preset large language model, which is then used to perform reasoning and analysis in conjunction with risk assessment results and customer profile information.

[0281] For example, when the risk level is low or medium, the large language model outputs a collection strategy. This collection strategy includes at least one of the following: collection method, communication focus, recommended collection script, suggested contact time, expected follow-up frequency, and precautions.

[0282] When the risk level is high, the big data model outputs a collection report. This report includes at least the collection strategy, customer status information (including status intent information and key objective information), historical communication information (including customer follow-up records), and risk assessment information (including comprehensive risk score and risk level). The collection strategy or report generated by the big data model is formatted and linked to overdue installment orders for subsequent push notifications.

[0283] Step S342: In response to a low risk level, the collection strategy is pushed to the sales staff so that the sales staff can perform collection operations according to the collection methods in the collection strategy.

[0284] For example, when the risk level of an overdue installment order is low, the salesperson responsible for the overdue installment order is identified as the target person, and the collection strategy corresponding to the overdue installment order is sent to the salesperson's corresponding business terminal (such as mobile phone or computer).

[0285] Sales personnel read the collection strategies, recommended collection scripts, and suggested contact times, and then execute at least one collection action based on the strategy, including telephone contact, instant messaging reminders, or SMS reminders. After completing the collection action, the result is recorded in the customer follow-up record.

[0286] Step S343: In response to the risk level being medium risk, the collection strategy is pushed to sales personnel and sales managers so that they can perform collection operations according to the collection methods in the collection strategy.

[0287] For example, when the risk level of an overdue installment order is medium risk, the sales personnel and sales supervisors responsible for the overdue installment order are identified as target personnel, and the same collection strategy is pushed to the business terminals (such as mobile phones or computers) of the sales personnel and sales supervisors respectively.

[0288] Sales staff execute collection actions according to the collection strategy and record the collection results. Sales supervisors monitor the collection progress of sales staff according to the collection strategy and provide collection guidance or adjust the collection plan based on the progress. The collection feedback from sales staff and sales supervisors is synchronously updated in the customer follow-up record.

[0289] Step S344: In response to a high-risk level, the collection report is pushed to sales personnel, sales supervisors, and the manual review platform. This allows sales personnel and sales supervisors to execute collection operations based on the collection strategy in the collection report, and allows risk control personnel to read the collection report in the manual review platform and review customer status information, historical communication information, and risk assessment information.

[0290] For example, when the risk level of an overdue installment order is high, the collection report will be pushed to the salesperson's business terminal, the sales supervisor's business terminal, and the manual review platform simultaneously.

[0291] Sales staff read the collection strategies in the collection report and execute collection actions accordingly. Sales supervisors, combining customer status information, historical communication information, and risk assessment information in the collection report, provide guidance to sales staff on their collection work and coordinate collection resources.

[0292] The manual review platform receives collection reports and generates corresponding manual review tasks. Risk control personnel log into the platform, retrieve customer status information, historical communication information, and risk assessment information from the collection reports, and review the collection strategies generated by the big data model based on these information. If the review is successful, the collection report is marked as approved, and the collection operation can continue according to the strategy outlined in the report. If the review fails, the risk control personnel revise or supplement the collection strategy, generating a revised strategy, which is then simultaneously pushed to sales personnel and sales managers for execution. The review results, revisions, and the final executed collection strategy are recorded in the customer follow-up log as historical data for subsequent risk assessments and big data model optimization.

[0293] In summary, the embodiments of this application have at least the following beneficial effects.

[0294] 1. Achieve more precise and comprehensive order risk assessment: By integrating eight indicators across four dimensions—orders, customers, sales personnel, and behavioral dynamics—a comprehensive customer profile is constructed, completely overcoming the one-sidedness of relying solely on the single indicator of "overdue days" for assessment. This makes the risk assessment results more comprehensive and closer to the actual business situation.

[0295] 2. Achieve dynamic, proactive, and self-evolving order risk warnings: 1) Enable risk assessment at the early stage of overdue payments or even before they occur, changing "post-event triggering" to "in-event warning"; 2) Through a weighted self-learning optimization and update mechanism, continuously and automatically optimize based on historical business results, achieving a fundamental shift from "static rules" to "dynamic intelligence" and maintaining long-term assessment accuracy.

[0296] 3. Achieving refined and intelligent order risk management strategies: Through continuous risk scoring and multi-level classification, the drawbacks of traditional "one-size-fits-all" grading are resolved. It integrates objective risk scores with customer profile information analyzed by a large language model, thereby matching or generating deeply personalized collection strategies for installment orders of different levels and statuses. This transforms passive response into proactive intelligent empowerment, greatly improving the efficiency and success rate of collection resource allocation.

[0297] 4. Achieve cognitive understanding of unstructured information and intelligent acceleration of processes: Through deep semantic analysis of customer follow-up records using a large language model, it can understand subjective information such as customer intent and emotions, and transform it into usable decision-making basis. This not only breaks through the limitations of traditional methods that only process structured data, but also automatically filters out extreme order cases such as "out of contact" before risk assessment, realizing intelligent process diversion and improving the overall processing efficiency of overdue installment orders.

[0298] This application provides a computer-readable storage medium storing processor-executable program instructions. When executed by the processor, the program instructions cause the processor to perform the steps of any possible implementation of the risk warning and collection method for installment payment equipment orders provided in this application, or the risk warning and collection method for installment payment equipment orders provided in this application.

[0299] In some embodiments, the storage medium may be a flash memory, a hard disk, an optical disk, a register, a magnetic surface memory, a removable disk, a CD-ROM, a random access memory (RAM), a read-only memory (ROM), an electrically programmable ROM, and an electrically erasable programmable ROM, or any other form of storage medium known in the art, or various devices including one or any combination of the above storage media.

[0300] In some embodiments, program 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.

[0301] As an example, program instructions may, but do not necessarily, correspond to files in the 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, or in a single file dedicated to the program in question, or in multiple collaborative files (such as files that store one or more modules, subroutines, or code sections).

[0302] As an example, program instructions can be deployed to execute on a computing device (including devices such as smart terminals and servers), or on multiple computing devices located in one location, or on multiple computing devices distributed across multiple locations and interconnected via a communication network. It is understood that all or part of the steps of the methods described in the embodiments provided in this application can be implemented directly using electronic hardware or processor-executable program instructions, or a combination of both.

[0303] Those skilled in the art will understand that the embodiments provided in this application are merely illustrative. The order in which the steps in the methods of the embodiments are written does not imply a strict execution order and does not constitute any limitation on the implementation process. The order can be adjusted, merged, and deleted according to actual needs. Modules or sub-modules, units or sub-units in the apparatus or system of the embodiments can be merged, divided, and deleted according to actual needs. For example, the division of units is only a logical functional division, and there may be other division methods in actual implementation. For another example, multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed.

[0304] Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented using software plus a general-purpose hardware platform, and of course, it can also be implemented using hardware. Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. This computer program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. It should be understood that the storage medium can be flash memory, hard disk, optical disk, register, magnetic surface memory, removable disk, CD-ROM, random access memory (RAM), read-only memory (ROM), electrically programmable ROM, and electrically erasable programmable ROM, etc.

[0305] It should be noted that the above embodiments are for illustrating the technical concept and features of this application, and are intended to enable those skilled in the art to understand the content of this application and implement it accordingly. They should not be construed as limiting the scope of protection of this application. Those skilled in the art can understand that all or part of the processes of the above embodiments can be implemented, modified according to the technical solutions described in the embodiments of this application, or equivalent substitutions can be made to some of the technical features. It is understood that these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application, and should be considered as equivalent changes and modifications made based on the embodiments of this application, all of which should fall within the scope of the claims of this application.

Claims

1. A method for risk warning and collection of payments for equipment orders with installment payment options, characterized in that, include: Obtain original order data, wherein the original order data includes overdue installment orders and order information, customer information, sales personnel information, repayment interaction information and customer follow-up records associated with the overdue installment orders; The customer follow-up records are input into a preset large language model. The large language model is used to extract state intention information and key objective information, and the state intention information and key objective information are integrated to obtain customer profile information. The state intention information is used to reflect the customer's repayment willingness and the degree of communication and cooperation, and the key objective information is used to reflect the customer's repayment ability and processing requests. Based on the customer profile information, a processing flag for the overdue installment order is generated, wherein the processing flag is used to indicate the processing method for the overdue installment order; In response to the processing flag being a risk assessment flag, a risk assessment is performed on the overdue installment order based on the original order data to obtain the comprehensive risk score and risk level of the overdue installment order. The risk assessment flag is used to indicate the assessment of the overdue installment order. Based on the comprehensive risk score, the risk level, and the customer profile information, a collection strategy for the overdue installment orders is generated so that the target personnel can perform collection operations according to the collection strategy.

2. The method according to claim 1, characterized in that, The order information includes the total order amount, overdue amount, overdue days, installment period, and product value indicator; the customer information includes the transaction activity index and historical overdue frequency; the sales personnel information includes the sales bad debt rate; and the repayment interaction information includes the collection response rate and overdue repayment amount. The step of conducting a risk assessment on the overdue installment orders based on the original order data to obtain a comprehensive risk score and risk level for the overdue installment orders includes: The first order score is determined based on the total order amount, the overdue amount, and the number of overdue days; The second order score is determined based on the number of installments. The third order score is determined based on the aforementioned commodity value indicator; The first customer score is determined based on the transaction activity index. The second customer score is determined based on the historical overdue frequency. The sales score is determined based on the aforementioned bad debt rate; The score for the first line is determined based on the aforementioned collection response rate; The score for the second line is determined based on the overdue payment amount. The first order score, the second order score, the third order score, the first customer score, the second customer score, the sales score, the first behavior score, and the second behavior score are weighted and combined to obtain the comprehensive risk score. The risk level of the overdue installment order is determined based on the comprehensive risk score.

3. The method according to claim 2, characterized in that, The step of determining the first order score based on the total order amount, the overdue amount, and the overdue days includes: The ratio of the overdue amount to the total order amount is determined as the overdue amount ratio; The time overdue coefficient is determined based on the sum of the overdue days and the first preset value; The product of the overdue amount ratio and the overdue time coefficient is determined as the overdue severity coefficient; The first order score is obtained based on the overdue severity coefficient and the preset first overdue coefficient, second overdue coefficient, third overdue coefficient, first increase coefficient, second increase coefficient, and third increase coefficient.

4. The method according to claim 3, characterized in that, The step of obtaining the first order score based on the overdue severity coefficient and preset first overdue coefficient, second overdue coefficient, third overdue coefficient, first increase coefficient, second increase coefficient, and third increase coefficient includes: In response to the fact that the overdue severity coefficient is greater than or equal to zero and less than the first overdue coefficient, the product of the first increase coefficient and the overdue severity coefficient is determined as the first order score; or, In response to the fact that the overdue severity coefficient is greater than or equal to the first overdue coefficient and less than the second overdue coefficient, the product of the first increase coefficient and the first overdue coefficient is determined as the first benchmark score; Multiply the second increase coefficient by the difference between the first coefficients to obtain the first coefficient product, where the first coefficient difference is the difference between the overdue severity coefficient and the first overdue coefficient. The sum of the product of the first benchmark score and the first coefficient is determined as the first order score; or, In response to the fact that the overdue severity coefficient is greater than or equal to the second overdue coefficient and less than the third overdue coefficient, the sum of the product of the first benchmark score and the second coefficient is determined as the second benchmark score, the product of the second coefficient is the product of the second increase coefficient and the difference of the overdue coefficient, and the difference of the overdue coefficient is the difference between the second overdue coefficient and the first overdue coefficient; Multiply the difference between the third increase coefficient and the second coefficient to obtain the product of the third coefficient, where the difference between the second coefficient is the difference between the overdue severity coefficient and the second overdue coefficient; The sum of the product of the second baseline score and the third coefficient is determined as the first order score; or, In response to the fact that the overdue severity coefficient is greater than or equal to the third overdue coefficient, a preset full score is determined as the first order score.

5. The method according to claim 2, characterized in that, The step of determining the second order score based on the installment period includes: A phased formula is constructed based on a preset base phased score, a reference phased score, a minimum number of phases, and a maximum number of phases, wherein the phased formula includes the number of phases; Substitute the number of installments into the installment formula to calculate the second order score.

6. The method according to claim 2, characterized in that, The customer information also includes the number of historical installment orders. The process of determining the second customer score based on the historical overdue frequency includes: In response to the historical installment order count being zero, a preset conservative overdue score is determined as the second customer score; or, In response to the historical overdue frequency being zero, a preset zero score is determined as the second customer score; or, In response to the historical delinquency frequency being greater than zero and less than or equal to a first frequency threshold, the product of a first frequency ratio and a preset first frequency score is determined as the second customer score, wherein the first frequency ratio is the ratio of the historical delinquency frequency to the first frequency threshold; or, In response to the historical overdue frequency being greater than the first frequency threshold and less than or equal to the second frequency threshold, the ratio of the first frequency difference to the reference frequency difference is determined as the second frequency ratio, where the first frequency difference is the difference between the historical overdue frequency and the first frequency threshold, and the reference frequency difference is the difference between the second frequency threshold and the first frequency threshold. The sum of the first frequency score and the first frequency product is determined as the second customer score, where the first frequency product is the product of the second frequency ratio and the preset second frequency score; or, In response to the historical overdue frequency being greater than the second frequency threshold, the ratio of the second frequency difference to the target frequency difference is determined as the third frequency ratio, where the second frequency difference is the difference between the historical overdue frequency and the second frequency threshold, and the target frequency difference is the difference between the standard frequency threshold 1 and the second frequency threshold. The minimum value between the second frequency product and the third frequency score is determined as the candidate score, wherein the second frequency product is the product of the third frequency ratio and the third frequency score, the third frequency score is the difference between a preset full score and a reference frequency score, and the reference frequency score is the sum of the first frequency score and the second frequency score; The sum of the reference frequency score and the candidate score is determined as the second customer score.

7. The method according to claim 2, characterized in that, The sales personnel information also includes the number of customers maintained, and the determination of the sales score based on the sales bad debt rate includes: In response to the number of maintained customers being less than a preset threshold, a preset conservative overdue score is determined as the sales score; or, In response to the number of maintained customers being greater than or equal to the preset quantity threshold and the sales bad debt rate being zero, a preset zero score is determined as the sales score; or, In response to the condition that the number of maintained customers is greater than or equal to the preset quantity threshold and the sales bad debt rate is greater than zero and less than or equal to the first bad debt rate threshold, the product of the first bad debt rate ratio and the preset first bad debt rate score is determined as the sales score, where the first bad debt rate ratio is the ratio of the sales bad debt rate to the first bad debt rate threshold; or, In response to the condition that the number of maintained customers is greater than or equal to the preset number threshold and the sales bad debt rate is greater than the first bad debt rate threshold and less than or equal to the second bad debt rate threshold, the ratio of the first bad debt rate difference to the reference bad debt rate difference is determined as the second bad debt rate ratio. The first bad debt rate difference is the difference between the sales bad debt rate and the first bad debt rate threshold, and the reference bad debt rate difference is the difference between the second bad debt rate threshold and the first bad debt rate threshold. The sales score is determined by summing the first bad debt rate score and the product of the first bad debt rate, where the first bad debt rate product is the product of the second bad debt rate ratio and a preset second bad debt rate score; or, In response to the fact that the number of maintained customers is greater than or equal to the preset number threshold and the sales bad debt rate is greater than the second bad debt rate threshold, the ratio of the second bad debt rate difference to the target bad debt rate difference is determined as the third bad debt rate ratio, wherein the second bad debt rate difference is the difference between the sales bad debt rate and the second bad debt rate threshold, and the target bad debt rate difference is the difference between the standard bad debt rate threshold 1 and the second bad debt rate threshold; The minimum value between the second bad debt rate product and the third bad debt rate score is determined as the undetermined score. The second bad debt rate product is the product of the third bad debt rate ratio and the third bad debt rate score. The third bad debt rate score is the difference between a preset full score and a reference bad debt rate score. The reference bad debt rate score is the sum of the first bad debt rate score and the second bad debt rate score. The sum of the reference bad debt rate score and the pending score is determined as the sales score.

8. The method according to claim 2, characterized in that, The determination of the second line score based on the overdue repayment amount includes: In response to the ratio of the overdue repayment amount to the overdue amount being greater than or equal to a preset ratio threshold, a preset zero score is determined as the score for the second row, wherein the overdue repayment amount is the actual repayment amount of the overdue installment order; In response to the ratio of the overdue repayment amount to the overdue amount being less than the ratio threshold, a preset full score is determined as the score for the second line.

9. The method according to claim 2, characterized in that, The step of weightedly aggregating the first order score, the second order score, the third order score, the first customer score, the second customer score, the sales score, the first behavior score, and the second behavior score to obtain the comprehensive risk score includes: Multiple target weights are obtained, including the first order weight corresponding to the first order score, the second order weight corresponding to the second order score, the third order weight corresponding to the third order score, the first customer weight corresponding to the first customer score, the second customer weight corresponding to the second customer score, the sales weight corresponding to the sales score, the first behavior weight corresponding to the first behavior score, and the second behavior weight corresponding to the second behavior score. The product of the first order weight and the first order score is determined as the first median score; The product of the second order weight and the second order score is determined as the second intermediate score; The product of the third order weight and the third order score is determined as the third intermediate score; The product of the first customer weight and the first customer score is determined as the fourth intermediate score; The product of the second customer weight and the second customer score is determined as the fifth intermediate score; The product of the sales weight and the sales score is determined as the sixth intermediate score; The product of the first behavior weight and the first behavior score is determined as the seventh intermediate score; The product of the second row weight and the second row score is determined as the eighth intermediate score; The comprehensive risk score is obtained by summing the first intermediate score, the second intermediate score, the third intermediate score, the fourth intermediate score, the fifth intermediate score, the sixth intermediate score, the seventh intermediate score, and the eighth intermediate score.

10. The method according to claim 9, characterized in that, The acquisition of multiple target weights includes: Obtain target order data, which includes multiple sample installment orders that have been overdue and completed. Each sample installment order is associated with order information, customer information, sales personnel information, and repayment interaction information. Based on multiple initial weights and the order information, customer information, sales personnel information, and repayment interaction information associated with the candidate installment orders, a risk assessment is performed on the candidate installment orders to obtain a comprehensive risk score for each candidate installment order. Each candidate installment order is any one of the multiple sample installment orders. The multiple initial weights include the first order initial weight corresponding to the first order score, the second order initial weight corresponding to the second order score, the third order initial weight corresponding to the third order score, the first customer initial weight corresponding to the first customer score, the second customer initial weight corresponding to the second customer score, the sales initial weight corresponding to the sales score, the first behavior initial weight corresponding to the first behavior score, and the second behavior initial weight corresponding to the second behavior score. Based on the order information, customer information, sales personnel information, and repayment interaction information associated with the candidate installment order, a result array corresponding to the candidate installment order is constructed. The result array includes a first value, a second value, and a third value. The first value is used to represent the processing result of the candidate installment order, the second value is used to represent the repayment time of the candidate installment order, and the third value is used to represent the collection cost-benefit ratio of the candidate installment order. Based on the result array and the comprehensive risk score of the candidate installment orders, the result category loss, payment time loss and cost-benefit ratio loss corresponding to the candidate installment orders are calculated; The comprehensive loss corresponding to the candidate installment order is obtained by weighted summing of the loss from the result classification, the loss from the payment time, and the loss from the cost-benefit ratio. The initial weights are iteratively updated based on the comprehensive loss corresponding to all the candidate installment orders until the comprehensive loss corresponding to all the candidate installment orders is minimized. The initial weights corresponding to the minimum overall loss for all candidate installment orders are determined as the final target weights.

11. The method according to claim 2, characterized in that, The process of determining the risk level of the overdue installment order based on the comprehensive risk score includes: In response to the comprehensive risk score being greater than or equal to zero and less than a preset first score threshold, a low-risk level is determined as the risk level of the overdue installment order; In response to the comprehensive risk score being greater than or equal to the first score threshold and less than a preset second score threshold, the medium risk level is determined as the risk level of the overdue installment order; In response to the overall risk score being greater than or equal to the second score threshold, a high-risk level is determined as the risk level of the overdue installment order.

12. The method according to any one of claims 1 to 11, characterized in that, The target personnel include one or more of sales personnel and sales supervisors. The step of generating a collection strategy for the overdue installment orders based on the comprehensive risk score, the risk level, and the customer profile information, so that the target personnel can execute collection operations according to the collection strategy, includes: The comprehensive risk score, the risk level, and the customer profile information are input into a preset big language model so that the big language model generates a collection strategy or a collection report. When the risk level is low or medium risk, the big language model generates a collection strategy. When the risk level is high risk, the big language model generates a collection report. The collection report includes the collection strategy, customer status information, historical communication information, and risk assessment information. In response to the risk level being low, the collection strategy is pushed to the sales personnel so that the sales personnel can perform collection operations according to the collection methods in the collection strategy; In response to the risk level being medium risk, the collection strategy is pushed to the sales personnel and the sales manager, so that the sales personnel and the sales manager can perform collection operations according to the collection methods in the collection strategy. In response to the risk level being high, the collection report is pushed to the sales personnel, the sales manager, and the manual review platform, so that the sales personnel and the sales manager can perform collection operations according to the collection strategy in the collection report, and the risk control personnel can read the collection report in the manual review platform and review the customer status information, the historical communication information, and the risk assessment information.

13. The method according to any one of claims 1 to 11, characterized in that, The method further includes: In response to the processing flag being a bad debt write-off flag, the overdue installment order is output to the manual review platform, enabling risk control personnel to perform write-off operations on the overdue installment order on the manual review platform. The bad debt write-off flag is used to indicate the write-off of the overdue installment order.

Images