Dynamic sorting method and device, computer equipment, readable storage medium and program product

Abstract

The invention relates to a dynamic sorting method and device, computer equipment, a computer readable storage medium and a computer program product. The method comprises the steps of obtaining to-be-processed business data of a plurality of target objects; key factors are extracted from the corresponding to-be-processed services, and comprehensive feature levels corresponding to the target objects are determined according to the key factors and a preset level sorting table; the preset grade sorting table is a corresponding relation table formed through multi-dimensional combination after discrete grading is carried out on the key factors based on historical data; the key factors comprise main key factors and other key factors; the main key factors are determined according to an optimization target; according to a statistical coefficient corresponding to the comprehensive feature level, dynamically correcting the main key factor to obtain a target key factor; and sorting the target objects by taking the target key factor as a primary sorting key and taking other key factors as secondary sorting keys to obtain a sorting result. By adopting the method, the sorting accuracy can be improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, and in particular to a dynamic sorting method and device, computer equipment, computer readable storage medium and computer program product. BACKGROUND

[0002] With the development of data analysis and intelligent decision-making technology, object sorting technology based on multi-factor scoring and priority judgment has emerged.

[0003] In related technologies, linear weighting methods based on a single or a small number of features are often used to sort objects. For example, a simple sum of willingness, success rate or numerical indicators according to fixed weights is obtained to get the comprehensive score of the target object and sort it accordingly. This kind of method is simple to implement, but since it ignores the mutual influence between multi-dimensional factors and the nonlinear relationship of historical data, the sorting result often has deviation and is difficult to truly reflect the comprehensive priority of the object. SUMMARY

[0004] Therefore, it is necessary to provide a dynamic sorting method, device, computer equipment, computer readable storage medium and computer program product capable of improving sorting accuracy in view of the above technical problems.

[0005] In a first aspect, the present application provides a dynamic sorting method, comprising:

[0006] obtaining business data to be processed of a plurality of target objects;

[0007] extracting key factors from each of the business data to be processed, and determining a comprehensive feature level of each of the target objects according to the key factors and a preset level sorting table; the preset level sorting table is a corresponding relationship table formed by multi-dimensional combination after the key factors are discretely graded based on historical data; the key factors include main key factors and other key factors; the main key factors are determined according to an optimization target;

[0008] dynamically correcting the main key factors according to a statistical coefficient corresponding to the comprehensive feature level to obtain target key factors;

[0009] sorting each of the target objects by taking the target key factors as a primary sorting key and the other key factors as a secondary sorting key to obtain a sorting result.

[0010] In one of the embodiments, the above determining a comprehensive feature level of each of the target objects according to the key factors and a preset level sorting table comprises:

[0011] Discretize each of the key factors, and match each of the key factors after discretization with the preset ranking table to determine the comprehensive feature level corresponding to each of the target objects.

[0012] In one of the embodiments, the other key factors include a willingness level; and the target key factor is obtained by correcting the main key factor according to the statistical coefficient corresponding to the comprehensive feature level, including:

[0013] The main key factor is corrected according to the statistical coefficient to obtain the corrected main key factor.

[0014] The target key factor is obtained based on the corrected main key factor, the willingness level, and a willingness influence index; the willingness influence index is determined by grid search of a willingness level influence weight in a preset parameter interval.

[0015] In one of the embodiments, the main key factor is corrected according to the statistical coefficient to obtain the corrected main key factor, including:

[0016] An interval coefficient is determined according to the main key factor.

[0017] The statistical coefficient is corrected according to the interval coefficient to obtain a correction coefficient.

[0018] The key factor is corrected based on the correction coefficient to obtain the corrected main key factor.

[0019] In one of the embodiments, the method further includes:

[0020] Time sequence feature data of each of the target objects is obtained.

[0021] The time sequence feature data is input into a pre-trained prediction model, and the time sequence feature data is predicted by the prediction model to obtain a predicted conversion time; the predicted conversion time is used to improve the business conversion rate.

[0022] In one of the embodiments, the method further includes:

[0023] A business conversion result is obtained; the business conversion result is obtained after performing a conversion process in a preset period.

[0024] The statistical coefficient and the willingness influence index are adjusted according to the business conversion result.

[0025] In a second aspect, the application further provides a dynamic ranking device, including:

[0026] An acquisition module is configured to acquire to-be-processed business data of a plurality of target objects;

[0027] A grade determination module is configured to extract key factors from the to-be-processed business data of each target object, and determine a comprehensive feature grade of each target object according to the key factors and a preset grade ranking table. The preset grade ranking table is a corresponding relationship table formed by multi-dimensional combination based on discrete grading of the key factors according to historical data. The key factors include main key factors and other key factors. The main key factors are determined according to an optimization target.

[0028] A correction module is configured to dynamically correct the main key factors according to a statistical coefficient corresponding to the comprehensive feature grade, to obtain target key factors.

[0029] A sorting module is configured to sort each target object according to the target key factors as a primary sorting key and the other key factors as a secondary sorting key, to obtain a sorting result.

[0030] In a third aspect, the present application further provides a computer device including a memory and a processor. The memory stores a computer program. The processor implements the steps of the method in any one of the above embodiments when executing the computer program.

[0031] In a fourth aspect, the present application further provides a computer readable storage medium having a computer program stored thereon. The computer program is executed by a processor to implement the steps of the method in any one of the above embodiments.

[0032] In a fifth aspect, the present application further provides a computer program product including a computer program. The computer program is executed by a processor to implement the steps of the method in any one of the above embodiments.

[0033] The above dynamic sorting method, device, computer device, computer readable storage medium and computer program product first extract a plurality of key factors from to-be-processed business data of each target object. Then, the key factors are discretely graded and processed, and a grade corresponding relationship table of multi-dimensional combination is constructed based on historical data to quantify the mutual influence relationship between different factors. Further, the main key factors are corrected according to a statistical coefficient corresponding to each comprehensive feature grade, so that the correction result can more accurately reflect the comprehensive influence law between multi-dimensional factors, thereby improving the representation ability of the key factors to the target result. Finally, the target key factors are taken as a primary sorting key, and the other key factors are taken as a secondary sorting key, to sort a plurality of target objects, thereby improving the accuracy of the sorting result under a global optimization target. BRIEF DESCRIPTION OF DRAWINGS

[0034] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0035] Figure 1 This is a diagram illustrating the application environment of the dynamic sorting method in one embodiment;

[0036] Figure 2 This is a flowchart illustrating a dynamic sorting method in one embodiment;

[0037] Figure 3 This is a schematic diagram illustrating the steps of a dynamic sorting method in a financial scenario in one embodiment;

[0038] Figure 4 This is a schematic diagram illustrating the steps of a dynamic sorting method in a resource scheduling scenario in one embodiment;

[0039] Figure 5 This is a structural block diagram of a dynamic sorting device in one embodiment;

[0040] Figure 6 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0041] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0042] The dynamic sorting method provided in this application embodiment can be applied to, for example, Figure 1In the application environment shown, terminal 102 communicates with server 104 via a network. A data storage system can store the data that server 104 needs to process. The data storage system can be integrated onto server 104 or placed on a cloud or other network server. Server 104 retrieves the pending business data of target objects from terminal 102, extracts key factors from the pending business data, and determines the comprehensive characteristic level of each target object based on the key factors and a preset ranking table. The preset ranking table is a correspondence table formed by multi-dimensional combination after discretely classifying key factors based on historical data. Key factors include main key factors and other key factors. The main key factors are determined according to the optimization objective. Based on the statistical coefficients corresponding to the comprehensive characteristic level, the main key factors are dynamically corrected to obtain the corrected main key factors. The corrected main key factors are used as the primary ranking key, and the other key factors are used as secondary ranking keys to sort each target object, obtaining the ranking result. The terminal 102 can be, but is not limited to, various personal computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices. IoT devices can include smart speakers, smart TVs, smart air conditioners, smart in-vehicle systems, and projection devices. Portable wearable devices can include smartwatches, smart bracelets, and head-mounted displays. Head-mounted displays can be virtual reality (VR) devices, augmented reality (AR) devices, and smart glasses. The server 104 can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services.

[0043] In one exemplary embodiment, such as Figure 2 As shown, a dynamic sorting method is provided, which can be applied to... Figure 1 Taking server 104 as an example, the explanation includes the following steps 202 to 206. Wherein:

[0044] Step 202: Obtain the pending business data of multiple target objects.

[0045] The target object refers to the entity object that needs to be sorted, predicted or processed in the current business process, such as users, accounts, devices or business units; the business data to be processed refers to the behavioral data of interaction with each target object, which may include, but is not limited to: object identification information, attribute information, behavior logs, time feature data, historical processing result data, etc.

[0046] Optionally, the business data to be processed can be obtained by the terminal device calling the interface of the server or data management system. For example, the terminal sends a data request to the data server by calling a preset application programming interface (API), and the server extracts the data record corresponding to the target object from the business database and returns it to the terminal; or, the terminal can receive the updated data of the target object from the real-time data stream through message queues, data buses, etc.

[0047] In other embodiments, the terminal can also periodically pull business data of multiple target objects from the distributed storage system through batch data synchronization tasks. For example, the terminal can pre-set the execution cycle of the data synchronization task (such as daily or hourly) and send synchronization instructions to each data node through the distributed task scheduling module; each node reads the target object business data fragments stored locally according to the instructions and preprocesses the data fragments, including field standardization, duplicate record filtering, and incremental update identification. Subsequently, the distributed storage system aggregates and merges the data fragments from different nodes to generate a dataset or batch data packet in a unified format.

[0048] Step 204: Extract key factors from each pending business, and determine the comprehensive characteristic level of each target object based on the key factors and the preset level ranking table. The preset level ranking table is a correspondence table formed by multi-dimensional combination after the key factors are discretely graded based on historical data. Key factors include main key factors and other key factors. The main key factors are determined according to the optimization objectives.

[0049] Key factors refer to the set of parameters used to characterize the business characteristics, behavioral characteristics, or state changes of the target object. Key factors can have a substantial impact on the processing results, ranking results, or prediction results of the target object. Key factors can include different types of data variables, such as numerical, categorical, or time-based variables.

[0050] For example, in the financial sector, key factors could be a customer's installment amount, installment willingness score, historical installment conversion rate, probability of phone call connection, number of days until the billing date, and historical click behavior.

[0051] For example, in the field of resource scheduling, key factors can be parameters that reflect resource allocation status, task execution efficiency, or load balancing. These may include, for instance, current resource occupancy rate, idle rate, task waiting time, scheduling success rate, task priority, transmission latency, node available capacity, energy consumption level, and historical scheduling feedback indicators. These key factors can characterize the dynamic changes of different resource units or task nodes, and have a significant impact on resource allocation strategies, task scheduling order, and overall system operating efficiency.

[0052] Optionally, each key factor can be discretized, and the discretized key factors can be matched with a preset ranking table to determine the comprehensive feature level of each target object.

[0053] The comprehensive feature level refers to the level label used to characterize the feature category of a target object under the combined effect of multiple key factors, reflecting the comprehensive performance of the target object in historical data. In determining the comprehensive feature level, each key factor is first discretized, and then the discretized key factors are matched with a preset level ranking table to obtain the comprehensive feature level corresponding to each target.

[0054] The preset ranking table is a correspondence table formed by multi-dimensional combination after discretely classifying the key factors based on historical data. The preset ranking table includes the combination relationships of each key factor under each level interval.

[0055] For example, assuming the key factors include key factor 1 and key factor 2, key factor 1 and key factor 2 are discretized to obtain several discrete levels (e.g., high, medium, low). A combined level table of the two is generated based on historical sample data, such as combinations like (high, high), (high, medium), and (medium, low). The server then matches the discretized result of the target object with a preset level ranking table to determine its corresponding comprehensive feature level.

[0056] For example, in the financial sector, key factors can include installment payment limits, willingness scores, and success rates. The server can categorize installment payment limits into ranges, classify willingness scores into high, medium, and low levels, and construct a pre-defined ranking table based on historical processing rate data. This allows for the matching and determination of comprehensive characteristic levels for different customers, which are then used for subsequent business priority calculations.

[0057] For example, in the field of resource scheduling, current resource occupancy rate, node idle rate, task waiting time, and task priority can be used as key factors. The system can discretize each factor, such as dividing the occupancy rate into three intervals: low, medium, and high, and the waiting time into three levels: short, medium, and long. Based on historical scheduling success rate or system response latency data, a preset level ranking table is constructed to determine the comprehensive characteristic level corresponding to each task node, which is used for subsequent task ranking and scheduling optimization.

[0058] Furthermore, the key factors include principal key factors and other key factors. The principal key factors are determined based on the optimization objective.

[0059] For example, in the field of resource scheduling, if the optimization objective is time, then the main key factor can be the task execution time; if the optimization objective is computing resources, then the main key factor can be the node CPU utilization.

[0060] For example, in the financial field, if the optimization objective is monetary value, then the primary factor could be the bill amount. When the optimization objective is risk minimization, the key factors could be the delinquency rate and the probability of abnormal transactions, etc.

[0061] Step 206: Based on the statistical coefficients corresponding to the comprehensive feature level, dynamically correct the main key factors to obtain the target key factors.

[0062] The key factors are interconnected and coupled, making it difficult to accurately reflect the synergistic effects among them when using only one major key factor for ranking or evaluation. Therefore, it is necessary to introduce statistical coefficients based on comprehensive characteristic levels to dynamically correct for the major key factors.

[0063] Optionally, statistical coefficients can be used to amplify or reduce the main key factors, for example, by multiplying the main key factors with the comprehensive characteristic level to obtain the target key factors.

[0064] Step 208: Sort each target object using the target key factor as the primary sorting key and other key factors as secondary sorting keys to obtain the sorting result.

[0065] Optionally, during the sorting process, there may be cases where the primary sort keys are the same. In this case, the secondary sort keys can be used for sorting.

[0066] Furthermore, the secondary sort keys can be sorted according to a preset method. The server can compare the values ​​of each secondary factor sequentially based on the importance of the secondary sort keys, or adopt a hierarchical sorting strategy: first compare the numerical values ​​of the primary sort keys, and if they are the same, continue to compare the secondary sort keys.

[0067] Optionally, different weight coefficients can be configured for the primary sorting key and the secondary sorting key. For example, the weight of the primary sorting key can be set to 0.7 to 0.9, and the weight of the secondary sorting key can be set to 0.1 to 0.3. The server can automatically adjust the weight ratio according to the business objectives and obtain the comprehensive sorting score through linear weighting or normalized weighting. This ensures that the main key factors play a dominant role while fully reflecting the auxiliary influence of other key factors on the overall sorting result, thus realizing a multi-dimensional comprehensive sorting mechanism that takes into account both primary and secondary factors.

[0068] In the aforementioned dynamic sorting method, firstly, multiple key factors are extracted from the business data to be processed for each target object. Then, these key factors are discretized and graded, and a multi-dimensional combination of grade correspondence tables is constructed based on historical data to quantify the mutual influence relationships between different factors. Further, the main key factors are corrected according to the statistical coefficients corresponding to each comprehensive characteristic grade, so that the correction results can more accurately reflect the comprehensive influence patterns between multi-dimensional factors, thereby improving the representational ability of key factors for the target results. Finally, the target key factors are used as the primary sorting key, and other key factors are used as secondary sorting keys to sort multiple target objects, improving the accuracy of the sorting results under the global optimization objective.

[0069] Furthermore, in one embodiment, the aforementioned other key factors include the willingness level; correspondingly, based on the statistical coefficients corresponding to the comprehensive feature level, the corresponding main key factors are corrected to obtain the corrected main key factors, including: correcting the corresponding main key factors based on the corresponding statistical coefficients to obtain the corrected main key factors; obtaining the corrected main key factors based on the corrected main key factors, the corresponding willingness level, and the willingness influence index; the corresponding willingness influence index is determined by performing a grid search on the willingness level influence weight within a preset parameter range.

[0070] In this context, willingness level refers to an indicator used to characterize the responsiveness of a target object in a specific behavior or task. In the financial field, willingness level can represent a customer's business response behavior characteristics over a historical period, such as whether they actively clicked on a business entry point, whether they had previously processed related business, response time, historical success rate, or installment payment willingness score. In the field of resource scheduling, willingness level can represent the response characteristics of task nodes or resource units in historical scheduling, such as node availability, task execution success rate, task reception frequency, average waiting time, or task response latency. These parameters can be used to characterize the node's willingness to cooperate and availability stability during the scheduling process.

[0071] Optionally, the corresponding key factors are corrected based on the corresponding statistical coefficients to obtain the corrected key factors, including: determining interval coefficients based on the corresponding key factors; correcting the corresponding statistical coefficients based on the corresponding interval coefficients to obtain correction coefficients; and correcting the corresponding key factors based on the corresponding correction coefficients to obtain the corrected key factors.

[0072] For example, in the financial field, one can analyze the fitted curve of the relationship between the installment amount and the installment processing rate α, divide the amount into three levels: "optimal", "second best" and "poor", and assign an interval coefficient to each level.

[0073] For example, in the field of resource scheduling, resource nodes can be classified based on historical distribution data of task execution results and resource utilization status. For instance, the server can divide nodes into three levels: "optimal," "second best," and "poor," based on indicators such as node resource utilization, task execution success rate, and average response time, and assign an interval coefficient to each level.

[0074] For example, the interval coefficient can be multiplied by the statistical coefficient to obtain the correction coefficient.

[0075] For example, the correction factor β = α * 0.8 for the poorest level; β = α * 1.0 for the second-best level; and β = α * 1.2 for the best level.

[0076] Then, the main key factors are corrected using correction coefficients to obtain the corrected main key factors.

[0077] Optionally, the correction coefficient can be multiplied by the key factor to obtain the corrected key factor. Taking the financial sector as an example, if the key factor is the actual installable amount, then the corrected key factor is the corrected installable amount, i.e., corrected installable amount = actual installable amount * β. Taking resource scheduling as an example, the key factor can be the task execution priority value. In this case, the corrected key factor can be the corrected task priority value, calculated as: corrected task priority value = original task priority value × β; where β is the resource scheduling coefficient, used to reflect the task execution success rate or scheduling reliability under different resource conditions.

[0078] In this way, by combining the main key factors with the correction coefficients, a dynamic weight balance can be achieved among multidimensional factors, comprehensively reflecting the interaction between features.

[0079] Finally, based on the corrected key factors, willingness levels, and willingness influence index, the target key factors are obtained.

[0080] Optionally, the server can calculate the willingness correction coefficient based on the willingness level and the willingness influence index, for example: Willingness correction coefficient = (willingness level)^γ.

[0081] The server can then perform a fusion calculation based on the stated intention correction coefficient and the corrected principal key factors to obtain the target key factors. For example, this can be achieved through multiplication or weighted fusion: Target key factor = Corrected principal key factors × Intention correction coefficient.

[0082] In this way, by adjusting the willingness level through the willingness influence index, the influence intensity of the willingness level in the overall calculation can be dynamically controlled.

[0083] For example, in the financial field, γ<1 serves to "downgrade the weight of high willingness and increase the importance of amount", preventing customers with extremely high willingness but very low amount from occupying too high priority.

[0084] For example, in a resource scheduling scenario, when the willingness influence index γ>1, the server will amplify the influence weight of the node response willingness level when calculating the key factors of the target, so that nodes with positive response and stable execution will obtain higher priority in the sorting.

[0085] The willingness influence index is determined by performing a grid search on the influence weight of willingness levels within a preset parameter range.

[0086] Optionally, the server sets multiple candidate parameter points within a preset parameter value range (e.g., γ∈[0.1, 2.0]), and calculates the model correction error or conversion accuracy under different parameters based on the correspondence between willingness levels and actual conversion results in historical business data. When a certain parameter value makes the matching degree between willingness level and conversion result the highest and the deviation the smallest, then that parameter is determined as the optimal willingness influence index γ.

[0087] Furthermore, the above method also includes: acquiring time-series feature data of each target object; inputting the time-series feature data into a pre-trained prediction model, using the prediction model to predict the time-series feature data, and obtaining the expected conversion time; the expected conversion time is used to improve the business conversion rate.

[0088] Time-series characteristic data refers to dynamic characteristic data that reflects the changing patterns of a target object over time, including but not limited to time-related behavioral records, state changes, or periodic indicators. Specifically, time-series characteristic data may include: the occurrence time of historical behaviors, periodic task records, resource usage curves over time, and business feedback results related to time windows, etc., used to characterize the trend features of the target object on the time axis.

[0089] Optionally, time-series feature data can be input into a pre-trained prediction model to obtain the estimated conversion time. The estimated conversion time refers to the moment when the target object is most likely to experience the target event or reach its optimal execution state within a future period. In the financial field, the estimated conversion time can be the point in time when a customer is most likely to accept business processing. In resource scheduling scenarios, the estimated conversion time can be the most suitable time for a node to receive tasks or migrate loads. For example, based on time-series features such as node CPU utilization, task execution cycle, response latency, and resource release trends, the optimal scheduling window for the node in the future period can be predicted so that tasks can be allocated when resources are idle and stable, thereby improving the overall system throughput and scheduling efficiency.

[0090] In this way, the server can combine time-series feature data to intelligently predict the timing of target behavior.

[0091] Furthermore, the above method also includes: obtaining business conversion results; using the business conversion results to perform conversion processing within a preset period; and adjusting statistical coefficients and intention impact index based on the business conversion results.

[0092] Optionally, after obtaining the estimated conversion time, the server will trigger or execute the corresponding conversion process within the corresponding estimated conversion time. For example, it may initiate business operations, resource scheduling, or task allocation on the target object at the optimal time, and record the actual execution result after the conversion process is completed. Here, the business conversion result refers to the actual response of the target object after the conversion process.

[0093] Optionally, if the business conversion result is lower than a preset threshold, the server can adaptively adjust the statistical coefficients and the willingness influence index according to the degree of deviation. For example, the server first calculates the difference between the actual conversion rate and the model-predicted conversion rate in the current period. When the difference exceeds the preset deviation threshold, a parameter update operation is triggered.

[0094] For example, when the actual conversion rate is lower than expected, the server can reduce the statistical coefficient corresponding to that combination, thereby reducing its weight in the calculation of the next cycle.

[0095] For example, when the actual conversion rate of a combination of key factors is higher than expected, the server increases the statistical coefficient corresponding to that combination to enhance its contribution to the prediction results.

[0096] In the above embodiments, statistical coefficients and willingness-influence indicators can be dynamically adjusted based on business conversion results to improve the accuracy of ranking.

[0097] In one exemplary embodiment, combined with Figure 3 , Figure 3 This is a schematic diagram illustrating the steps of a dynamic sorting method in a financial scenario in one embodiment.

[0098] Step S302: Obtain the user's pending business data.

[0099] For example, firstly, the system extracts historical and real-time data related to the target direct sales business (such as transaction installment business) from data sources such as business databases and user behavior log libraries through data interfaces.

[0100] Step S304: Extract key factors from the data to be processed for each user.

[0101] By using statistical analysis, correlation analysis and other methods, we identified and determined several key factors that affect the conversion of business processing.

[0102] Key factors include, but are not limited to: Customer's installment payment amount: The customer's current credit limit or bill amount available for installment payments; Installment payment willingness score: A quantitative score representing the likelihood of a customer applying for installment payments, generated by a predictive model built using machine learning algorithms such as XGBoost based on the customer's historical behavior data; Historical installment payment conversion rate: The percentage of customers who have successfully applied for installment payments through marketing in the past; Call connection probability: A quantitative score representing the likelihood of a customer answering unsolicited marketing calls, generated by a predictive model built using the LightGBM model based on the customer's historical call answering behavior data; Number of days until the billing date: A dynamic, time-sensitive factor reflecting the customer's current billing cycle stage; Historical click behavior: Whether the customer has recently (e.g., within a week) clicked on installment-related links in the app or SMS (tracked events), which is a strong real-time intention signal.

[0103] Step 306: Discretize the key factors and determine the user's comprehensive characteristic level based on the discretized key factors.

[0104] For example, the willingness to pay in installments is divided into three levels: high, medium, and low; the probability of a phone call being connected is divided into 10 levels: 1-10; the number of days between the billing date and the installment payment rate is finely divided into 31 levels based on the fitted curve of the interval and the installment payment rate, and then further divided into 7 main levels to simplify the model; and whether the installment payment tracking point is clicked within a week is divided into two levels: "yes" and "no".

[0105] Next, the discretized key factors are matched with a preset ranking table. This table combines different levels of all factors, theoretically creating 1860 (3*10*31*2) different user profile tiers. The comprehensive characteristic level of each user profile tier is the average installment payment rate (denoted as α) of the user group within a historical period (e.g., the previous 3 months) for each characteristic combination tier. This α value serves as the basic conversion probability for this type of customer group.

[0106] Step 308: Based on the statistical coefficients corresponding to the comprehensive feature level, dynamically correct the main key factors to obtain the target key factors.

[0107] In this embodiment, the optimization objective is to increase the transaction amount, rather than simply increasing the number of conversions. Therefore, the transaction amount is considered the primary key factor, and adjustments are made accordingly.

[0108] Amount Tier Adjustment: First, analyze the fitted curve of the relationship between the installment amount and the installment processing rate α, dividing the amount into three tiers: "optimal," "second best," and "poor." Assign an installment processing coefficient β to each tier (poor tier β = α * 0.8; second best tier β = α * 1.0; optimal tier β = α * 1.2). Then, adjust the customer's actual installment amount using β, resulting in the adjusted installment amount = actual installment amount * β. This step aims to balance "amount" and "conversion probability," avoiding blindly pursuing high-amount customers with no intention to convert.

[0109] Expected Installment Amount Calculation: To obtain a final ranking index that comprehensively reflects customer value and conversion potential, a grid search method is introduced to find an optimal index γ (commonly 0.7), and the customer's expected installment amount is calculated using the following formula:

[0110] Expected installment amount = Adjusted installment amount * (Installment willingness level)^γ

[0111] This formula uses an exponent γ to adjust the weight of "intention level". γ < 1 has the effect of "downgrading the weight of high intention and increasing the importance of amount", preventing customers with extremely high intention but low amount from occupying too high a priority.

[0112] Step 310: Sort each target object using the target key factor as the primary sorting key and other key factors as secondary sorting keys to obtain the sorting result.

[0113] The system sorts users in descending order using the calculated expected installment amount as the primary sorting key. For customers with similar or identical expected installment amounts, secondary sorting keys are used, including installment willingness score, probability of phone call connection, and whether installment payment tracking was used. This results in a comprehensive user priority list.

[0114] Step 312: Obtain the time-series characteristic data of each user and calculate the expected conversion time based on the time-series characteristic data.

[0115] The time-series feature data refers to the model's prediction of the expected conversion time for each user in the user priority list based on time-series data such as the customer's billing date, repayment date, and historical call count. This allows staff to execute conversion processes within the expected conversion time. This may include triggering customer contact tasks, sending business notifications, pushing processing reminders, or generating business interaction instructions within the system during the expected conversion time, thus prompting customers to complete business conversion at the optimal time.

[0116] Step 314: Obtain business conversion results and adjust the statistical system and intention impact indicators based on the business conversion results.

[0117] In this embodiment, the results of the conversion process can be monitored through a real-time monitoring dashboard, so as to adjust the statistical system and the intention-influence indicators in real time.

[0118] Optionally, the server can recalculate each update α and β based on the daily conversion results.

[0119] In the above embodiments, by comprehensively modeling and dynamically correcting multidimensional business data of users, the integration and optimization of subjective willingness factors and objective business indicators are realized; combined with the predicted conversion time based on time-series feature data, the conversion process can be triggered at the optimal time, thereby improving the timeliness and success rate of business execution; at the same time, by introducing a feedback mechanism for business conversion results, the statistical coefficients and willingness influence index are dynamically adjusted, enabling the model to have self-learning and adaptive capabilities, continuously optimizing the ranking and prediction results, and ultimately significantly improving the business conversion rate and system resource utilization efficiency.

[0120] In one exemplary embodiment, combined with Figure 4 , Figure 4 This is a schematic diagram illustrating the steps of a dynamic sorting method in a resource scheduling scenario in one embodiment.

[0121] Step 402: Obtain the pending business data for each scheduling object.

[0122] The server extracts historical and real-time data related to multiple scheduling objects (compute nodes / containers / service instances) from the task queue, monitoring platform, and operation and maintenance database, including: node identifier, CPU / memory / bandwidth usage, queue length, average response latency, task completion rate, failure rate, number of retries, and the time of the most recent maintenance / restart.

[0123] Step 404: Extract key factors from the data to be processed corresponding to each scheduling object.

[0124] The server uses statistical and correlation analysis to identify key factors affecting scheduling success rate and overall efficiency. These key factors include, but are not limited to: CPU availability, memory availability, bandwidth availability, I / O idle time, average response latency, historical task success rate, task waiting time, and queue backlog rate.

[0125] Step 406: Discretize the key factors and determine the comprehensive characteristic level of the scheduling object based on the discretized key factors.

[0126] Key factors are discretized into levels; for example, CPU utilization can be divided into high, medium, and low levels. Then, the discretized key factors are matched with a preset level ranking table. This preset ranking table is a correspondence table formed by multi-dimensional combinations of each discrete level based on historical data. Each combination corresponds to a historical statistical coefficient α.

[0127] For example, after discretizing each key factor, it is matched with a preset ranking table to obtain a comprehensive feature level.

[0128] Step 408: Based on the statistical coefficients corresponding to the comprehensive feature level, dynamically correct the main key factors to obtain the target key factors.

[0129] For example, the intervals can be divided into "optimal intervals", "second-best intervals", and "poor intervals", with each interval corresponding to a different interval coefficient. For instance, the interval coefficient for a poor interval could be 0.8, for a second-best interval could be 1.0, and for an optimal interval could be 1.2.

[0130] The server can then use interval coefficients to correct the statistical coefficients, for example, by multiplying the interval coefficients by the statistical coefficients to obtain corrected coefficients. Based on these corrected coefficients, the key factors are then weighted and corrected to obtain the target key factors.

[0131] For example, in a resource scheduling scenario, the main key factors can be the amount of available resources of a node or the priority of task execution. If a node is in the optimal range, the server sets its range coefficient to 1.2 to amplify the impact of resource status on the sorting; if a node is in the poor range, the range coefficient is set to 0.8 to reduce the priority of the node in the scheduling sorting.

[0132] Step 410: Sort each target object using the target key factor as the primary sorting key and other key factors as secondary sorting keys to obtain the sorting result.

[0133] In this step, the sorting process can be referred to in any of the above embodiments, and will not be repeated here.

[0134] Step 412: Obtain the timing characteristic data of each scheduling object, and calculate the expected conversion time based on the timing characteristic data.

[0135] The server acquires the time-series characteristic data of each scheduled object and calculates the estimated transition time based on this data. This time-series characteristic data may include trends in node CPU utilization, task cycle patterns, peak and off-peak periods, response latency curves, and resource release cycles. The server can input this time-series characteristic data into a pre-trained time prediction model to predict the most suitable time for each scheduled object to execute its task within a future time period, i.e., the estimated transition time.

[0136] The server can prioritize task allocation or trigger scheduling operations within the expected conversion time, thereby improving the success rate of task execution and the overall system throughput efficiency.

[0137] Step 414: Obtain business conversion results and adjust the statistical system and intention impact indicators based on the business conversion results.

[0138] After a task is executed or scheduled, the server obtains the business conversion results and dynamically adjusts the statistical coefficients and the willingness impact index based on these results. The business conversion results include actual feedback data such as task completion rate, latency deviation, anomaly rate, or resource utilization compliance rate within a preset period.

[0139] When the business conversion result is lower than the preset threshold, the server can reduce the impact of the willingness level on the ranking weight by lowering the statistical coefficient corresponding to the poor performance factor or appropriately lowering the willingness influence index. When the business conversion result is better than expected, the server can increase the statistical coefficient corresponding to the high performance node and appropriately raise the willingness influence index to enhance the positive effect of the willingness feature.

[0140] In the above embodiments, intelligent sorting and optimal timing prediction of resource scheduling objects are achieved through multi-dimensional factor fusion and dynamic correction mechanisms. This method can adaptively adjust weights based on the resource status, historical performance, and response willingness of nodes, and form a self-learning closed loop by combining time-series feature prediction and feedback optimization, thereby significantly improving the accuracy of task allocation and system resource utilization efficiency, and ensuring the stability and real-time performance of the scheduling process.

[0141] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0142] Based on the same inventive concept, this application also provides a dynamic sorting apparatus for implementing the dynamic sorting method described above. The solution provided by this apparatus is similar to the implementation described in the above method; therefore, the specific limitations in one or more embodiments of the dynamic sorting apparatus provided below can be found in the limitations of the dynamic sorting method described above, and will not be repeated here.

[0143] In one exemplary embodiment, such as Figure 5As shown, a dynamic sorting device is provided, including: an acquisition module, a level determination module, a correction module, and a sorting module, wherein:

[0144] The acquisition module 100 is used to acquire pending business data from multiple target objects.

[0145] The grade determination module 200 is used to extract key factors from each pending business and determine the comprehensive characteristic grade of each target object based on the key factors and the preset grade ranking table. The preset grade ranking table is a correspondence table formed by multi-dimensional combination after the key factors are discretely graded based on historical data. The key factors include the main key factors and other key factors. The main key factors are determined according to the optimization objectives.

[0146] The correction module 300 is used to dynamically correct the main key factors based on the statistical coefficients corresponding to the comprehensive feature level, so as to obtain the target key factors.

[0147] The sorting module 400 is used to sort each target object by using the target key factor as the primary sorting key and other key factors as secondary sorting keys, and obtain the sorting result.

[0148] In one embodiment, the grade determination module 200 includes:

[0149] The level matching unit is used to discretize each key factor and match each discretized key factor with a preset level ranking table to determine the comprehensive feature level corresponding to each target object.

[0150] In one embodiment, the correction module 300 includes:

[0151] The factor correction unit is used to correct the main key factors based on statistical coefficients to obtain the corrected main key factors.

[0152] The calculation unit is used to obtain the target key factors based on the corrected main key factors, willingness level, and willingness influence index; the willingness influence index is determined by performing a grid search on the influence weight of willingness level within a preset parameter range.

[0153] In one embodiment, the above-mentioned factor correction unit includes:

[0154] The coefficient determination sub-unit is used to determine the interval coefficients based on the main key factors.

[0155] The coefficient correction subunit is used to correct the statistical coefficients based on the interval coefficients to obtain the correction coefficients.

[0156] The factor correction subunit is used to correct the main key factors based on the correction coefficients to obtain the corrected main key factors.

[0157] In one embodiment, the above-mentioned apparatus further includes:

[0158] The data acquisition module is used to acquire the time-series characteristic data of each target object.

[0159] The prediction module is used to input time-series feature data into a pre-trained prediction model, and then use the prediction model to predict the time-series feature data to obtain the expected conversion time; the expected conversion time is used to improve the business conversion rate.

[0160] In one embodiment, the above-mentioned apparatus further includes:

[0161] The result acquisition module is used to obtain business conversion results; the business conversion results are obtained after performing conversion processing within a preset period.

[0162] The adjustment module is used to adjust the statistical coefficients and the willingness impact index based on the business conversion results.

[0163] Each module in the aforementioned dynamic sorting device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of a computer device in hardware form or independent of it, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0164] In one exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 6 As shown, this computer device includes a processor, memory, input / output interfaces (I / O), and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores pending business data. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external terminals via a network connection. When executed by the processor, the computer program implements a dynamic sorting method.

[0165] Those skilled in the art will understand that Figure 6The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0166] In one exemplary embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the method in any of the above embodiments.

[0167] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps of the method in any of the above embodiments.

[0168] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps of the method in any of the embodiments.

[0169] 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. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0170] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0171] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A dynamic sorting method, characterized in that, The method includes: Retrieve pending business data from multiple target objects; Key factors are extracted from each of the pending business operations. Based on the key factors and a preset ranking table, the comprehensive characteristic level of each target object is determined. The preset ranking table is a correspondence table formed by multi-dimensional combination after the key factors are discretely graded based on historical data. The key factors include main key factors and other key factors. The main key factors are determined according to the optimization objective. Based on the statistical coefficients corresponding to the comprehensive feature levels, the main key factors are dynamically corrected to obtain the target key factors; The target objects are sorted by using the target key factors as the primary sorting key and the other key factors as secondary sorting keys, to obtain the sorting result.

2. The method according to claim 1, characterized in that, The step of determining the comprehensive feature level corresponding to each target object based on the key factors and the preset level ranking table includes: Each of the key factors is discretized, and the discretized key factors are matched with the preset level ranking table to determine the comprehensive feature level corresponding to each target object.

3. The method according to claim 1, characterized in that, The other key factors include the willingness level; the step of correcting the main key factors based on the statistical coefficients corresponding to the comprehensive feature level to obtain the target key factors includes: Based on the statistical coefficients, the key factors are corrected to obtain the corrected key factors; Based on the corrected key factors, the willingness level, and the willingness influence index, the target key factors are obtained; the willingness influence index is determined by performing a grid search on the influence weight of the willingness level within a preset parameter range.

4. The method according to claim 3, characterized in that, The step of correcting the key factors based on the statistical coefficients to obtain the corrected key factors includes: Based on the aforementioned key factors, determine the interval coefficients; The statistical coefficients are corrected based on the interval coefficients to obtain the correction coefficients; Based on the correction coefficient, the main key factors are corrected to obtain the corrected main key factors.

5. The method according to claim 3, characterized in that, The method further includes: Obtain the temporal feature data of each of the target objects; The time-series feature data is input into a pre-trained prediction model, and the prediction model is used to predict the time-series feature data to obtain the expected conversion time; the expected conversion time is used to improve the business conversion rate.

6. The method according to claim 5, characterized in that, The method further includes: Obtain business conversion results; the business conversion results are obtained after performing conversion processing within a preset period; Based on the business conversion results, adjust the statistical coefficients and the willingness impact index.

7. A dynamic sorting device, characterized in that, The device includes: The acquisition module is used to acquire pending business data from multiple target objects. The grade determination module is used to extract key factors from each of the pending business operations, and determine the comprehensive characteristic grade of each target object based on the key factors and a preset grade ranking table. The preset grade ranking table is a correspondence table formed by multi-dimensional combination after discretely classifying the key factors based on historical data. The key factors include main key factors and other key factors. The main key factors are determined according to the optimization objective. The correction module is used to dynamically correct the main key factors based on the statistical coefficients corresponding to the comprehensive feature level, so as to obtain the target key factors; The sorting module is used to sort the target objects by using the target key factors as the primary sorting key and the other key factors as secondary sorting keys, so as to obtain the sorting result.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

Images