A method, device and system for screening accounting data processing methods

Abstract

The embodiments of this specification disclose a method, device and system for screening account data processing methods, the method comprising receiving an account data processing method screening request, the account data processing method screening request comprising an account identifier; obtaining a first probability value, a guarantee type and an overdue type corresponding to the account identifier; the first probability value is used to characterize the probability that the user cannot repay the amount of resources involved in the account data within a preset time; determining the collection queue corresponding to the current account data according to the first probability value, the guarantee type and the overdue type corresponding to the account identifier; determining the second probability value that the current account data may become a bad debt at the corresponding overdue time by using the downward migration rate of the collection queue to which the current account data belongs at each overdue time; and determining the processing method corresponding to the current account data at the corresponding overdue time according to the second probability value. Using each embodiment, the efficiency of account data processing can be improved, and the risk of resource loss of financial institutions can be reduced.

Description

Technical Field

[0001] The present specification relates to the field of computer data processing technology, and in particular, to a method, device and system for screening accounting data processing methods. Background Art

[0002] With the rapid growth of loan business, the workload of business processing has also increased significantly. However, the current business processing system of financial institutions lacks a more systematic and effective processing method configuration scheme for the account data corresponding to such business, and usually directly relies on manual personal experience to process account data. However, manual experience is overly dependent on the experience and personal quality of the formulator, and cannot effectively guarantee the risk of resource loss that may be caused when using the corresponding processing method to process account data. Therefore, there is an urgent need for a more accurate and effective account data processing method configuration scheme to improve the efficiency of account data processing and reduce the risk of resource loss of financial institutions. Summary of the invention

[0003] The purpose of the embodiments of this specification is to provide a method, device and system for screening accounting data processing methods, which can improve the efficiency of accounting data processing and reduce the risk of resource loss of financial institutions.

[0004] This specification provides a method, device and system for screening accounting data processing methods, which are implemented in the following ways:

[0005] A method for screening an account data processing method, applied to a server, comprises: receiving an account data processing method screening request, the account data processing method screening request comprising an account identifier; the account identifier is used to identify the current account data corresponding to the account data processing method screening request. Obtaining a first probability value, guarantee type, and overdue type corresponding to the account identifier; the first probability value is used to characterize the probability that a user cannot repay the amount of resources involved in the account data within a preset time. Determining a collection queue corresponding to the current account data according to the first probability value, guarantee type, and overdue type corresponding to the account identifier; the collection queue is pre-divided according to the first probability value, guarantee type, and overdue type of the account data. The second probability value that the current account data may become a bad debt at the corresponding overdue time is determined by using the downgrade rate of the collection queue to which the current account data belongs at each overdue time; the downgrade rate is the ratio of the number of bad debts in the collection queue to the number of accounts in the collection queue at the corresponding overdue time; the bad debt includes account data that cannot be recovered within a preset time; the overdue time includes a time value that exceeds the specified repayment time of the account data. According to the second probability value, the processing method corresponding to the current account data at the corresponding overdue time is selected from the account data processing method set.

[0006] In some other embodiments of the method provided in this specification, the first probability value corresponding to the account identifier is determined in the following manner: obtain the credit principal overdue characteristic data, credit interest overdue characteristic data and guarantee overdue characteristic data of the user corresponding to the current account data. Input the credit principal overdue characteristic data into the credit principal risk identification model to obtain a first risk identification score. Input the credit interest overdue characteristic data into the credit interest risk identification model to obtain a second risk identification score. Input the guarantee overdue characteristic data into the guarantee risk identification model to obtain a third risk identification score. Determine the first probability value corresponding to the account identifier based on the first risk identification score, the second risk identification score and the third risk identification score.

[0007] In some other embodiments of the method provided in this specification, the method of selecting the processing method corresponding to the current account data at the corresponding overdue time from the account data processing method set according to the second probability value includes: obtaining the latest collection time of the collection queue to which the current account data belongs under different processing methods; wherein the latest collection time is determined according to the distribution characteristics of the account data recovery rate of each collection queue relative to the overdue time. The processing method adopted by the current account data at each overdue time is determined according to the latest collection time and the second probability value.

[0008] In some other embodiments of the method provided in this specification, the processing method includes a first collection channel, a second collection channel and a third collection channel; wherein the probability of recovering account data of the first collection channel, the second collection channel and the third collection channel decreases in sequence.

[0009] In some other embodiments of the method provided in this specification, the method of determining the processing method adopted by the current account data at each overdue time according to the latest collection time and the second probability value includes: judging whether the current account data meets the first collection condition, and the first collection condition includes: the second probability value of the current account data becoming a bad debt at the specified overdue time is greater than or equal to the first collection threshold, or the specified overdue time is greater than or equal to the first latest collection time corresponding to the collection queue to which the current account data belongs under the first collection channel. If yes, the first collection channel is executed starting from the specified overdue time.

[0010] In some other embodiments of the method provided in this specification, the method also includes: if the designated overdue time is the first overdue day of the current account data, then executing collection channels other than the first collection channel from the designated overdue time to N days after the designated overdue time; and executing the first collection channel starting from N days after the designated overdue time, where N is greater than or equal to 1.

[0011] In some other embodiments of the method provided in this specification, the method of determining the processing method adopted by the current account data at each overdue time according to the latest collection time and the second probability value includes: judging whether the current account data meets the second collection condition, the second collection condition includes: the second probability value of the current account data becoming a bad debt at the specified overdue time is greater than or equal to the second collection threshold and less than the first collection threshold, or the specified overdue time is greater than or equal to the second latest collection time corresponding to the collection queue to which the current account data belongs under the second collection channel; wherein the first collection threshold is greater than the second collection threshold, and the first latest collection time is greater than the second latest collection time. If yes, the second collection channel is executed from the specified overdue time to the first latest collection time.

[0012] In some other embodiments of the method provided in this specification, when the current account data satisfies neither the first collection condition nor the second collection condition, the third collection channel is executed.

[0013] In some other embodiments of the method provided in this specification, the first collection threshold is determined in the following manner: Calculate the difference between the executable upper limit of the first collection account corresponding to the first collection channel and the amount of accounts to be collected covered by the latest collection time corresponding to the first collection channel to obtain the first account number. Obtain the maximum downward migration rate of each account data settled before the latest collection time corresponding to the first collection channel, and the maximum downward migration rate is equal to the downward migration rate of the collection queue where each settled account data is located on the corresponding settlement date. Sort the account data settled before the latest collection time corresponding to the first collection channel in descending order according to the maximum downward migration rate, and use the maximum downward migration rate of the account data ranked in the first account number as the first collection threshold.

[0014] In some other embodiments of the method provided in this specification, the second collection threshold is determined in the following manner: Calculate the difference between the executable upper limit of the second collection account corresponding to the second collection channel and the amount of accounts to be collected covered by the latest collection time corresponding to the second collection channel to obtain the second account number. Obtain the maximum downward migration rate of each account data settled before the latest collection time corresponding to the second collection channel; wherein the maximum downward migration rate is equal to the downward migration rate of the collection queue where the settled account data is located on the settlement date of the corresponding account data. Sort the account data settled before the latest collection time corresponding to the second collection channel in descending order according to the maximum downward migration rate, and use the maximum downward migration rate of the account data ranked in the second account number as the second collection threshold.

[0015] On the other hand, the embodiments of this specification also provide an account data processing method screening device, which is applied to a server and includes: a data receiving module, which is used to receive an account data processing method screening request, and the account data processing method screening request includes an account identifier; the account identifier is used to identify the current account data corresponding to the account data processing method screening request. A data acquisition module, which is used to obtain a first probability value, a guarantee type, and an overdue type corresponding to the account identifier; the first probability value is used to characterize the probability that the user cannot repay the amount of resources involved in the account data within a preset time. A collection queue determination module, which is used to determine the collection queue corresponding to the current account data according to the first probability value, the guarantee type, and the overdue type corresponding to the account identifier; the collection queue is pre-divided according to the first probability value of the account data and the guarantee type and overdue type. A probability value determination module is used to determine a second probability value that the current account data may become a bad debt at the corresponding overdue time by using the downward migration rate of the collection queue to which the current account data belongs at each overdue time; the downward migration rate is the ratio of the number of bad debts in the collection queue to the number of accounts in the collection queue at the corresponding overdue time; the bad debt includes account data that cannot be recovered within a preset time; the overdue time includes a time value that exceeds the specified repayment time of the account data. A processing method screening module is used to screen out the processing method corresponding to the current account data at the corresponding overdue time from the account data processing method set according to the second probability value.

[0016] On the other hand, an embodiment of the present specification also provides a system for screening accounting data processing methods, the system comprising at least one processor and a memory for storing processor executable instructions, the instructions implementing the steps of the method described in any one or more of the above embodiments when executed by the processor.

[0017] The method, device and system for screening the processing method of account data provided by one or more embodiments of this specification can analyze the account data from the perspective of guarantee method, overdue type and the first probability value that the user corresponding to the overdue account cannot repay the account data in time, and divide the collection queues corresponding to each account data. Correspondingly, each collection queue has similar basic data characteristics and risk characteristics. Then, the second probability value of the account data that has not been settled during the corresponding overdue period can be quantified using the distribution of the bad debt ratio of each collection queue with the overdue time. Afterwards, the second probability value can be used to configure the processing method of the account data that has not been settled during the corresponding overdue period, thereby achieving differentiated configuration of the processing methods of account data with different risks, greatly improving the collection recovery rate of account data and the reasonable allocation of limited collection resources, improving the efficiency of account data processing, and reducing the risk of resource loss of financial institutions. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the embodiments of this specification or the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art description. Obviously, the drawings described below are only some embodiments recorded in this specification. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor. In the drawings:

[0019] Figure 1 A flowchart of an embodiment of a method for screening an accounting data processing method provided in this specification;

[0020] Figure 2 A schematic diagram of the division of collection queues in an embodiment provided in this specification;

[0021] Figure 3 A schematic diagram of the distribution of the recovery rate of accounting data over time in another embodiment provided in this specification;

[0022] Figure 4 A schematic diagram of an optimization method for processing in another embodiment provided in this specification;

[0023] Figure 5 A schematic diagram of a processing method evaluation system in another embodiment provided in this specification;

[0024] Figure 6 A schematic diagram of a first probability value determination process in another embodiment provided in this specification;

[0025] Figure 7 A schematic diagram of the distribution of migration rates under different guarantee methods in another embodiment provided in this specification;

[0026] Figure 8 A schematic diagram of the distribution of downward migration rates under different overdue types in another embodiment provided in this specification;

[0027] Fig. 9 A schematic diagram of the sample size and downward migration rate distribution of a credit account data sample in another embodiment provided in this specification;

[0028] Fig.10 A schematic diagram of the sample size and downward migration rate distribution of guarantee account data samples in another embodiment provided in this specification;

[0029] Fig.11 A schematic diagram of a credit principal overdue characteristic data type in another embodiment provided in this specification;

[0030] Fig.12 A schematic diagram of correlation coefficients of credit principal overdue feature data types in another embodiment provided in this specification;

[0031] Fig.13 A schematic diagram of the discrimination degree of a credit principal risk identification model constructed based on training samples in another embodiment provided in this specification;

[0032] Fig.14 A schematic diagram of the discrimination degree of a credit principal risk identification model optimized based on a test sample in another embodiment provided in this specification;

[0033] Fig.15 A schematic diagram of the discrimination degree of a credit principal risk identification model based on a verification sample in another embodiment provided in this specification;

[0034] Fig.16 A schematic diagram of the module structure of a device for screening accounting data processing methods provided in this specification;

[0035] Fig.17 A schematic structural diagram of a server according to an exemplary embodiment of the present specification. DETAILED DESCRIPTION

[0036] In order to enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in one or more embodiments of this specification will be clearly and completely described below in conjunction with the drawings in one or more embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of the specification, not all of the embodiments. Based on one or more embodiments of the specification, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the embodiments of this specification.

[0037] Accounting data can refer to the transaction data generated by users' transactions of financial products such as loans in financial institutions. The transaction types, guarantee methods, overdue types and corresponding users corresponding to the accounting data are quite different. The data is huge and the data analysis is complicated and tedious. How to accurately and efficiently determine the processing method corresponding to the accounting data, improve the efficiency of accounting data processing, and reduce the risk of resource loss of financial institutions has become a technical problem that needs to be solved urgently.

[0038] In the embodiments of this specification, the collection queues corresponding to each account data are divided by analyzing the guarantee method, overdue type, and the probability that the corresponding user cannot repay the amount of resources involved in the account data within the preset time. Correspondingly, each collection queue has similar basic data characteristics and risk characteristics. Then, the second probability value of the account data that has not been settled during the corresponding overdue period can be quantified using the distribution of the bad debt ratio of each collection queue with the overdue time. Afterwards, the second probability value can be used to configure the processing method of the account data that has not been settled during the corresponding overdue period, thereby achieving differentiated configuration of the processing method of account data with different risks, greatly reducing the risk of resource loss of financial institutions, and improving the efficiency of account data processing.

[0039] The processing method may include collection channels, execution time of each collection channel, collection words, etc. The processing method may be configured in advance according to business scenarios and collection resources, etc. The collection channels may include, for example, telephone collection, SMS collection, and door-to-door collection, etc. The telephone collection may also include, for example, robot telephone collection, manual telephone collection, etc.

[0040] In an example of an application scenario of this specification, the server of the business processing system can send a determined processing method to the terminal device of the business personnel who perform the collection, so that the terminal device can display it, and the corresponding business personnel can collect the debts according to the displayed processing method. Alternatively, the server can also send the determined processing method to the terminal device that automatically performs the collection, so that the terminal device completes the collection according to the processing method. Alternatively, the server can also perform different operations based on different collection channels. For example, when determining SMS collection, the server can send the processing method to the terminal device that performs SMS collection, so that the terminal device sends a collection SMS to the terminal device of the corresponding user. When determining door-to-door or telephone collection, the server sends the processing method to the terminal device of the business personnel who perform the collection, so that it can be displayed on the terminal device, and the corresponding business personnel can collect the debts according to the displayed processing method.

[0041] Figure 1It is a flowchart of an embodiment of the method for screening the accounting data processing method provided in this specification. Although this specification provides the method operation steps or device structure shown in the following embodiments or drawings, more or fewer operation steps or module units may be included in the method or device based on routine or no creative labor. In the steps or structures where there is no necessary causal relationship logically, the execution order of these steps or the module structure of the device is not limited to the execution order or module structure shown in the embodiments or drawings of this specification. When the method or module structure is applied in an actual device, server or terminal product, it can be executed sequentially or in parallel according to the method or module structure shown in the embodiment or drawings (for example, a parallel processor or multi-threaded processing environment, or even a distributed processing, server cluster implementation environment).

[0042] A specific implementation example is Figure 1 As shown, in one embodiment of the method for screening the accounting data processing mode provided in this specification, the method can be applied to the server, and the method can include the following steps:

[0043] S20: receiving an account data processing method screening request, wherein the account data processing method screening request includes an account identifier; the account identifier is used to identify current account data corresponding to the account data processing method screening request.

[0044] The server may receive a request for filtering an account data processing method. The request for filtering an account data processing method may include an account identifier. The account identifier may be used to identify the current account data corresponding to the request for filtering an account data processing method. The current account data may refer to any account data to be analyzed.

[0045] S22: Obtain a first probability value, guarantee type, and overdue type corresponding to the account identifier; the first probability value is used to represent the probability that the user cannot repay the amount of resources involved in the account data within a preset time.

[0046] The server may obtain the first probability value, guarantee type, and overdue type corresponding to the account identifier.

[0047] The guarantee methods may include credit guarantee, surety guarantee, mortgage guarantee, etc. The overdue types may include principal overdue and interest overdue.

[0048] The first probability value may be used to represent the probability that the user cannot repay the amount of resources involved in the account data within a preset time.

[0049] The server can obtain the basic data of the current account data from the business system of the financial institution. The basic information may include the account identifier of the account data, the user identifier corresponding to the account identifier, the guarantee type, the overdue type, the overdue days, etc. The server can obtain the basic data of the account data from the business system based on a preset time interval, or can obtain the basic data of the account data from the business system at any time.

[0050] The first probability value corresponding to the current account data can be pre-evaluated and determined based on the business data, repayment behavior data, account data, etc. of the user corresponding to the current account data. The first probability value corresponding to the current account data can be retrieved from the business system by the server at the same time as or after acquiring the basic data, based on the user identifier of the current account data, and then determined based on the retrieved data. Alternatively, each first probability value can be pre-evaluated by a server or other device, and then, after acquiring the basic data, the server retrieves the corresponding first probability value based on the user identifier. Of course, the above data acquisition method is only an example of a preferred embodiment, and can be configured according to the actual business scenario during specific implementation, and is not limited here.

[0051] S24: Determine the collection queue corresponding to the current account data according to the first probability value, guarantee type and overdue type corresponding to the account identifier; the collection queue is pre-divided according to the first probability value, guarantee type and overdue type of the account data.

[0052] The server may determine the collection queue to which the current account data belongs according to the first probability value corresponding to the current account data and the guarantee type and overdue type of the current account data.

[0053] In some implementations, the server may pre-divide the collection queues by analyzing the first probability values ​​corresponding to a large number of account data samples and the guarantee types and overdue types of the account data samples. For example, the server divides the collection queues based on a pre-configured collection queue allocation algorithm in combination with one or more of the above factors. For example, the collection queues may be divided using algorithms such as decision trees and GBDT. By pre-dividing the collection queues in the above manner, each collection queue may have similar basic business characteristics and risk levels, so that different processing methods can be configured in a targeted manner based on the collection queues, so that collection resources can be used more reasonably, the collection recovery rate of account data can be improved, and the user experience can be improved.

[0054] In other implementations, in addition to guarantee type and overdue type, the server can further divide the collection queue based on features such as whether the account data is used in high-risk industries, whether the account corresponding to the account data is on a blacklist such as involved in litigation or has credit problems, and whether the type of financial product corresponding to the account data is more likely to become a bad debt. Figure 2 As shown, the embodiment of this specification finally divides into 11 collection queues, of which there are 3 collection queues corresponding to credit principal and credit interest, 2 collection queues corresponding to guarantee principal, and 3 collection queues corresponding to guarantee interest. Figure 2 It can be seen that the basic business characteristics, risk levels and bad debt ratios (bad rates) of each collection queue are significantly different.

[0055] S26: Determine a second probability value that the current account data may become a bad debt at the corresponding overdue time by using the downward migration rate of the collection queue to which the current account data belongs at each overdue time; the downward migration rate is the ratio of the number of bad debts in the collection queue to the number of accounts in the collection queue at the corresponding overdue time; the bad debt includes account data that cannot be recovered within a preset time; the overdue time includes a time value that exceeds the specified repayment time of the account data.

[0056] The server can use the downward migration rate of the collection queue to which the current account data belongs at each overdue time to determine the second probability value that the current account data may become a bad debt at the corresponding overdue time. The downward migration rate can be the ratio of the number of bad debts in the collection queue to the number of accounts in the collection queue at the corresponding overdue time. The overdue time may include a time value that exceeds the specified repayment time of the account data. The bad debt may include account data that has not been completed within a preset time. It should be noted that the specific limit of the bad debt can be set according to the account processing situation of the financial institution, and the above description is only a preferred description and is not a direct limitation.

[0057] Then, the server can differentiate the second probability value of the current account data that may become a bad debt according to the different overdue times. The possibility of account data becoming a bad debt varies greatly with different overdue times. By differentiating the second probability value of account data becoming a bad debt according to the overdue time, the processing methods that should be adopted for account data at different overdue times can be further refined, the detailed configuration of the processing methods can be improved, and collection resources can be reasonably allocated to improve the user experience while ensuring the recovery rate.

[0058] The downward migration rate of the collection queue to which the current account data belongs at each overdue time can be used as the second probability value of the current account data becoming a bad debt at the corresponding overdue time, and the downward migration rate is the ratio of the number of bad debts in the collection queue to the number of accounts in the collection queue at the corresponding overdue time. The account data of the same collection queue have similar basic information characteristics and risk characteristics. The bad debt ratio of the collection queue at a certain overdue time is used to quantify the second probability value of the account data that has not been settled at the overdue time may become a bad debt at the overdue time. This can more efficiently and accurately realize the distribution characteristic analysis of the second probability value of the account data that may become a bad debt at each overdue time, and improve the accuracy of the processing method configuration.

[0059] S28: Filtering out the processing method corresponding to the current account data at the corresponding overdue time from the account data processing method set according to the second probability value.

[0060] The server may pre-build a set of account data processing methods, which may include processing methods corresponding to account data at different overdue times and different second probability values. In practical applications, the server may filter out the processing method used by the current account data at the corresponding overdue time from the set of account data processing methods according to the second probability value of the current account data becoming a bad debt at each overdue time.

[0061] For example, the greater the second probability value of a current account data becoming a bad debt at a certain overdue time, the higher the cost but higher probability of recovery can be configured for the current account data at the corresponding overdue time, such as door-to-door collection by business personnel. The smaller the second probability value of a current account data becoming a bad debt at a certain overdue time, the lower the cost but lower probability of recovery can be configured for the current account data at the corresponding overdue time, such as SMS collection.

[0062] The second probability value division threshold corresponding to each processing method can be configured based on a large amount of sample data analysis and actual collection resources. The server can be pre-configured with processing methods corresponding to each second probability value threshold range. After determining the second probability value of the current account data becoming a bad debt, the corresponding processing method can be determined based on the comparison between the second probability value and the threshold.

[0063] The scheme of the above embodiment is based on the risk of the user corresponding to the account data not being able to repay the account in time, the guarantee type of the account data itself, and the overdue type. The risk of the account data becoming a bad debt is initially quantified, and then the risk of the account data becoming a bad debt changes with the overdue time. The collection resources are configured differently. Through the above configuration method, the account data with higher risk can be implemented as soon as possible. The processing method with a higher recovery probability is implemented to improve the recovery rate; while the account data with lower risk can be implemented first because whether to use the processing method and what processing method to use does not affect its recovery probability. The processing method with a lower recovery probability but lower cost can be implemented first, thereby further realizing the reasonable allocation of limited collection resources on the basis of ensuring the recovery rate. However, with the change of overdue time, the risk of the account data with a lower risk that has not been repaid is gradually increasing. Further combining the risk change law of the account data with the overdue time to perform differentiated allocation of collection resources, the recovery rate of the account data can be improved on the basis of ensuring the reasonable allocation of limited collection resources, thereby reducing the risk of resource loss of financial institutions on the basis of improving the efficiency of account data processing.

[0064] In some other embodiments provided in this specification, the method may further include:

[0065] S210: Obtain the latest collection time of the collection queue to which the current account data belongs under different processing methods; wherein the latest collection time is determined according to the distribution characteristics of the account data recovery rate of each collection queue relative to the overdue time.

[0066] S212: Determine the processing method used for the current account data at each overdue time according to the latest collection time and the second probability value.

[0067] The server may pre-configure the latest collection time for each collection queue under different processing methods. In some embodiments, the server may determine the latest collection time for each collection queue under different processing methods based on the distribution characteristics of the account data recovery rate of each collection queue relative to the overdue time. The latest collection time corresponding to different processing methods is used to indicate that the corresponding processing method needs to be executed from the latest collection time to ensure the collection recovery rate.

[0068] First, we analyze the recovery rate of account data. We can use the calculation formula: Recovery rate on T overdue day = (Amount of account data on T overdue day – Amount of overdue accounts on T+4 overdue days) / Amount of recoverable account data on T overdue day, where Amount of recoverable account data on T overdue day = Amount of account data on T overdue day – Amount of bad debts (bad debts are estimated based on the amount of account data that has not been paid off after a preset time period).

[0069] like Figure 3 As shown, Figure 3It shows the recovery rate distribution of account data that is overdue for 1 to 30 days. Based on the three commonly used processing methods: SMS collection, telephone collection and door-to-door collection, the following conclusions can be drawn by analyzing the recovery rate characteristics of account data:

[0070] a. Assuming there is no collection action, the account data recovery rate under natural repayment should show a trend of first decreasing and then fluctuating steadily.

[0071] b. The recovery rate of credit interest account data (this type of account data is usually most sensitive to collection actions) shows a significant increase in 8-13 days and 22-29 days. Through analysis of actual operations, it can be seen that business personnel are usually accustomed to taking door-to-door collection actions during these two periods, which also shows that door-to-door collection actions during the corresponding periods have a more obvious recovery effect on interest account data.

[0072] c. The recovery rate of credit principal account data increases slightly after 23 days of overdue, but the increase is small, which indicates that the principal account data is difficult to recover quickly at this stage, and the principal account data should be collected as early as possible.

[0073] d. The recovery rate of overdue guarantee interest increases after 20 days, which is similar to the trend of overdue credit interest, but the amplitude is smaller. Business personnel will also conduct door-to-door collection of guarantee account data at this stage, but the intensity is less than that of credit account data.

[0074] e. Due to the small number of cases, the recovery rate of guarantee principal account data is not very representative. The recovery rate of guarantee interest can be used as a reference for analysis.

[0075] According to the analysis of the distribution pattern of recovery rate with overdue time, the latest door-to-door collection time for credit principal account data, credit interest account data, guarantee principal account data, and guarantee interest account data can be preliminarily determined. The latest telephone collection time can be determined based on the door-to-door collection time, such as one or two days before the latest door-to-door collection time. Of course, in actual application scenarios, after designing a processing method based on the above-mentioned latest collection time and the second probability value, the recovery rate distribution characteristics after collection based on the designed processing method can be further analyzed to adjust the latest collection time so that the setting of the latest collection time is more in line with the actual application scenario.

[0076] By further determining the latest collection time of each processing method in each collection queue, and designing and adjusting the processing method in combination with the latest collection time and the second probability value of the account data, the configuration of the processing method can be further made more in line with the actual application scenario, thereby improving the rationality of resource allocation and the recovery rate.

[0077] In some implementations, the processing method may include a first collection channel, a second collection channel, and a third collection channel. The probability of recovering the account data of the first collection channel, the second collection channel, and the third collection channel decreases in sequence. The first collection channel may be, for example, door-to-door collection, the second collection channel may be, for example, telephone collection, and the third door-to-door channel may be, for example, SMS collection. Based on the above three processing methods, the processing method used for the account data at each overdue time may be further accurately configured according to the latest collection time and the second probability value.

[0078] In some embodiments, the server may determine whether the current account data satisfies a first collection condition, and the first collection condition may include: a second probability value of the current account data becoming a bad debt at a specified overdue time is greater than or equal to a first collection threshold, or the specified overdue time is greater than or equal to a first latest collection time corresponding to the collection queue to which the current account data belongs under the first collection channel. If yes, the first collection channel is executed starting from the specified overdue time.

[0079] In other embodiments, if the designated overdue time is the first day of overdueness of the current account data, collection channels other than the first collection channel are executed from the designated overdue time to N days after the designated overdue time; and the first collection channel is executed starting from N days after the designated overdue time, where N is greater than or equal to 1.

[0080] In some other embodiments, the server may determine whether the current account data satisfies the second collection condition, and the second collection condition may include: the second probability value of the current account data becoming a bad debt at the specified overdue time is greater than or equal to the second collection threshold and less than the first collection threshold, or the specified overdue time is greater than or equal to the second latest collection time corresponding to the collection queue to which the current account data belongs under the second collection channel. Wherein, the first collection threshold is greater than the second collection threshold, and the first latest collection time is greater than the second latest collection time; if so, the second collection channel is executed from the specified overdue time to the first latest collection time.

[0081] In some other embodiments, the server may execute the third collection channel when the current account data satisfies neither the first collection condition nor the second collection condition. Alternatively, the server may execute the third collection channel for auxiliary collection while executing the first collection channel or the second collection channel. Alternatively, the server may continue to execute the third collection channel after the account data enters the bad debt stage.

[0082] The first collection threshold and the second collection threshold can be configured according to the actual business scenario. The first latest collection time and the second latest collection time can be determined according to the latest collection time in the above embodiment. It should be noted that the first latest collection time and the second latest collection time corresponding to the first collection channel and the second collection channel under different collection queues may be different. The specific values ​​of the first latest collection time and the second latest collection time need to be determined in combination with the collection queue to which the current account data belongs.

[0083] By quantifying the resource allocation of account data in the above way, the second probability value of account data and the processing method can be matched more accurately, the utilization efficiency of collection resources can be improved, and the collection effect can be improved. At the same time, the collection cost and complaint rate can be more reasonably controlled, and the user experience can be improved.

[0084] In some embodiments, the first collection threshold may be determined in the following manner:

[0085] Calculate the difference between the executable upper limit of the first collection account corresponding to the first collection channel and the amount of the to-be-collected accounts covered by the latest collection time corresponding to the first collection channel to obtain the first account number;

[0086] Obtaining a maximum downward migration rate of each account data settled before the latest collection time corresponding to the first collection channel, wherein the maximum downward migration rate is equal to the downward migration rate of the collection queue where each settled account data is located on the corresponding settlement date;

[0087] The account data settled before the latest collection time corresponding to the first collection channel are sorted in descending order according to the maximum downward migration rate, and the maximum downward migration rate of the account data ranked first is used as the first collection threshold.

[0088] In some other embodiments, the second collection threshold may be determined in the following manner:

[0089] Calculate the difference between the executable upper limit of the second collection account corresponding to the second collection channel and the amount of the to-be-collected accounts covered by the latest collection time corresponding to the second collection channel to obtain the second account number;

[0090] Obtaining the maximum downward migration rate of each account data that is settled before the latest collection time corresponding to the second collection channel; wherein the maximum downward migration rate of the account data is equal to the downward migration rate of the collection queue where the settled account data is located on the settlement date of the corresponding account data;

[0091] The account data settled before the latest collection time corresponding to the second collection channel are sorted in descending order according to the maximum downward migration rate, and the maximum downward migration rate of the account data ranked second is used as the second collection threshold.

[0092] Taking the first collection channel as door-to-door collection, the second collection channel as telephone collection, and the third collection channel as SMS collection as examples, the above-mentioned first collection threshold and second collection threshold determination method are explained. A two-dimensional table can be formed by the collection queue-overdue days to calculate the number of accounts to be collected and the downgrade rate of each collection queue at each overdue day. The executable upper limit of collection accounts for door-to-door collection and telephone collection can be estimated based on the current total amount of collection accounts data and available collection resources to ensure that the higher-cost processing method can complete the collection task and try not to be idle.

[0093] At the time of door-to-door collection, considering the collection pressure of business personnel, each collection queue can pre-configure the latest door-to-door collection time corresponding to the door-to-door collection. Assume that the sum of the amount of account data to be collected at the latest door-to-door collection time is X, and the upper limit of the first collection account that can be executed is Y. Because Y is the maximum executable task volume, and X is the amount of tasks to be collected after the latest door-to-door collection time is limited, the value of Y is usually greater than X, which means that there is still collection capacity left for door-to-door collection. This part of the remaining collection capacity can be further redistributed according to the second probability value, so that the account data with a higher second probability value can be collected as early as possible before the latest door-to-door collection time.

[0094] The account data settled by each collection queue before the latest door-to-door collection time can be screened out, and the maximum downward migration rate can be calculated. The maximum downward migration rate is equal to the downward migration rate of the collection queue where the settled account data is located on the corresponding settlement date. The account data settled before the latest door-to-door collection time are sorted in order from high to low according to the maximum downward migration rate. The maximum downward migration rate ranked at the front Z position is taken as the door-to-door collection threshold, that is, the first collection threshold. Among them, the value of Z is the difference between Y and X. That is, before the latest door-to-door collection time, the remaining collection capacity will be further reasonably redistributed according to the second probability value of the account data, further improving the early collection of account data with higher risks and improving the collection recovery rate.

[0095] For telephone collection, the same method as door-to-door collection is used to determine the second collection threshold corresponding to the telephone collection, which will not be elaborated here.

[0096] The account data of the credit principal high and medium risk queues usually have a higher second probability value and are often configured to visit the customer on the first day of overdue payment. However, business personnel may not be able to reach the user in time. To increase the reach rate, they can arrange collection calls on the first day and visit the customer on the second day.

[0097] In telephone collection, each collection queue can be configured with multiple outbound call time points. Based on business experience and consideration of user experience, the collection frequency of each account data can be set according to the second probability value of different collection queues to match the executable collection limit of telephone collection.

[0098] After determining the door-to-door collection time and telephone collection time, the rules for sending SMS collection can be further configured based on business experience and data analysis. For example, SMS collection can be executed before the telephone collection time, or SMS collection can be used as an auxiliary collection method while collecting by phone or door-to-door collection. Of course, during the door-to-door collection process, if there is a surplus in telephone collection, telephone collection can also be used as an auxiliary collection method. In some implementation methods, the specific content of telephone collection and SMS collection can also be configured, and the collection content can be used as one of the information of the processing method.

[0099] The configuration of the processing method based on the above method can make the configuration of the processing method clearer and more specific, and various collection methods can perform their respective functions, accurately match user risks, and improve the rationality of collection resource allocation and recovery rate. At the same time, SMS collection is used first for low-risk account data to reduce interference with users, which can further improve the user experience.

[0100] In other embodiments, the processing method can also be monitored to optimize and adjust the processing method. Among the indicators of strategy monitoring, the recovery rate (collection rate), migration rate, and bad debt rate can be selected as core evaluation indicators. The collection cost, complaint rate, and response rate are selected as process indicators. The recovery efficiency is used as a comprehensive indicator, where recovery efficiency = recovery amount / collection cost. A strategy monitoring report is designed for longitudinal monitoring of the strategy operation process, which is convenient for observing fluctuations in strategy effectiveness and making timely adjustments and optimizations. It can also be used for management assessment of the collection team. Use such Figure 4 The champion challenger scheme shown in the figure iteratively optimizes the strategy and establishes Figure 5 The RNC policy evaluation system shown is used to optimize the processing method so that the processing method is more in line with the actual application scenario.

[0101] like Figure 6 As shown, in some other embodiments, the first probability value corresponding to the current account data can be determined in the following manner:

[0102] S220: Acquire credit principal overdue characteristic data, credit interest overdue characteristic data, and guarantee overdue characteristic data of the user corresponding to the account identifier;

[0103] S222: inputting the credit principal overdue characteristic data into a credit principal risk identification model to obtain a first risk identification score;

[0104] S224: inputting the credit interest overdue characteristic data into a credit interest risk identification model to obtain a second risk identification score;

[0105] S226: Inputting the guarantee overdue characteristic data into a guarantee risk identification model to obtain a third risk identification score;

[0106] S228: Determine a first probability value corresponding to the account identifier according to the first risk identification score, the second risk identification score, and the third risk identification score.

[0107] The server may analyze a large number of account data samples in advance and divide the account data samples into good samples and bad samples. The good samples may refer to account data that completes the account data settlement within a predetermined time. The bad samples may refer to account data that does not complete the account data settlement within a predetermined time. The ratio of the good samples to the bad samples may be set according to the actual business scenario.

[0108] The server can determine the feature data type and extraction method based on the guarantee method, overdue type and the migration distribution characteristics of account data from one overdue stage to another overdue stage, so that the risk of account data becoming bad debt can be more accurately estimated based on the extracted feature data.

[0109] The overdue stages can be divided according to the number of overdue days, and the overdue characteristics of account data can be analyzed according to the overdue stages to predict the rules of account data in different overdue stages becoming bad debts in the future. For example, the overdue stages can be divided into eight stages from M0 to M7+. No overdue is M0, 1 to 29 days overdue is M1, 30 to 59 days overdue is M2, and so on. Overdue for more than 180 days is defined as M7+.

[0110] In some embodiments, the downward migration rate distribution of a large number of account data samples under different guarantee methods and overdue types can be analyzed to determine the type of characteristic data and the extraction method. The downward migration rate can characterize the migration distribution characteristics of account data from a certain overdue stage to another overdue stage, and thus predict future bad debt losses at different overdue stages.

[0111] Figure 7 The following is a diagram showing the distribution of the downgrade rate under different guarantee methods as a function of the number of days overdue. Figure 7 As shown in the figure, the credit category has a higher downgrade rate, which is significantly different from other guarantee methods and can be considered as a separate category. Although the guarantee characteristics of the guarantee are similar to those of the credit category in actual operation, its downgrade rate after 30 days is closer to that of the mortgage category. At the same time, since the sample size of the guarantee category is significantly smaller than that of the credit category, it is not suitable to be included in the credit model. In this embodiment, the guarantee category is also included in the guarantee category. Figure 7As shown in the figure, the downward migration rates of credit and guarantee types tend to stabilize when they are 30 days overdue and 60 days overdue, respectively. That is, the performance period is 30 days after the credit type is overdue, and the performance period is 60 days after the guarantee type is overdue. Therefore, the account data that is paid off within 30 / 60 days after the overdue period can be defined as a good sample; the account data that is overdue for more than 30 / 60 days is defined as a bad debt.

[0112] Figure 8 The following is a schematic diagram showing the distribution of the down-migration rate under different overdue types as a function of overdue time. Figure 8 As shown in the figure, there is a significant difference in the downward migration rate between principal overdue and interest overdue, and the ratio of the downward migration rate between principal overdue and interest overdue is nearly 7:1. The overdue type also has a great impact on the analysis of risk identification. Risk analysis can be conducted based on different overdue types to further improve the accuracy of risk analysis. Due to the insufficient sample size of guarantee type, the guarantee model does not distinguish between overdue types for the time being, but it can be included as a variable in the model during the model construction process.

[0113] We can further analyze the distribution of the sample size of credit and guarantee account data samples over time and the distribution of the migration rate over time, and determine the data time distribution period of the account data samples required for model construction, so that the extracted feature data can more accurately and comprehensively reflect the risk of the account data becoming bad debts. Fig. 9 and Fig.10 As shown, Fig. 9 It represents the distribution of the sample size of credit account data samples over time and the distribution of the downward migration rate over time. Fig.10 The sample size of the guarantee account data sample is distributed over time, as well as the distribution of the downward migration rate over time. The downward migration rate of principal overdue fluctuated steadily after November 2018, while the downward migration rate of interest overdue was stable after September 2018. Considering that the sample is selected for a full natural year, the account data from June 2018 to June 2019 can be selected as the modeling sample. Similarly, in the guarantee category, the modeling sample can also be selected based on the sample size and the fluctuation of the downward migration rate. Considering that the sample size of the guarantee category is small, the account data from January 2017 to July 2019 can be selected as the modeling sample.

[0114] After determining the classification and data extraction cycle of feature data extraction, business data can be filtered through aging data, credit data, behavioral data, enterprise information, business owner characteristics, etc., and an enterprise label system can be constructed through six major blocks: basic enterprise characteristics, enterprise repayment characteristics, enterprise overdue characteristics, enterprise behavior characteristics, enterprise asset characteristics, and enterprise risk characteristics, in order to filter feature data types. More than 1,600 feature data types are determined directly based on the above business data and enterprise label system. There are many feature data types, and there are a lot of redundant and noisy data. The feature data types can be further filtered through the following steps:

[0115] a. Delete the feature data types with missing value ratio>=0.8 or maximum single value ratio>=0.8 or feature data types with value types>=30 (character type);

[0116] b. Use the "Best-KS" optimal binning algorithm to bin feature data types;

[0117] c. Calculate the characteristic data types iv and psi, and delete iv<0.1 or psi>0.1 or if they do not match;

[0118] Then, you can further filter feature data types based on business explanation and variable relevance:

[0119] d. WOE mapping, calculate the Pearson correlation coefficient between feature data types, and only retain one of the feature data type groups with correlation > 0.6;

[0120] e. Algorithm selection & business experience adjustment.

[0121] Among them, IV (Information Value) is used to measure the predictive ability of variables. Suppose in a classification problem, there are two categories of target variables: Y1 and Y2. For an individual A to be predicted, to determine whether A belongs to Y1 or Y2, certain information is required. Assume that the total amount of information is I. And this required information is contained in all independent variables C1, C2, C3, ..., Cn. Then, for one of the variables Ci, the more information it contains, the greater its contribution to judging whether A belongs to Y1 or Y2, the greater the information value of Ci, the greater the IV of Ci, and the more Ci should be considered as one of the candidate variables for model entry.

[0122] PSI (Population Stability Index) is an indicator used to measure the deviation between the predicted value and the actual value of the model.

[0123] WOE (Weight of Evidence) is a form of encoding the original independent variable. To encode a variable with WOE, you need to first bin the variable. After binning, for the i-th group, the calculation formula of WOE is as follows:

[0124]

[0125] Where: p y1 is the proportion of bad samples to all bad samples, p y0 The ratio of good samples to all good samples; B is the total number of bad samples, B iis the number of bad samples corresponding to variable i, G is the total number of good samples, G i is the number of good samples corresponding to variable i. The larger the WOE, the more likely the samples in this group are bad samples, and the smaller the WOE, the smaller the possibility of bad samples in this group.

[0126] The Pearson correlation coefficient (Px,y) of two continuous variables (X,Y) is equal to the covariance between them cov(X,Y) divided by the product of their respective standard deviations (σX,σY). The value of the coefficient is always between -1.0 and 1.0. Variables close to 0 are called uncorrelated, and those close to 1 or -1 are called strongly correlated.

[0127] After the above feature screening, there are 12 types of credit principal overdue feature data, 10 types of credit interest overdue feature data, and 15 types of guarantee overdue feature data. Fig.11 This is a diagram showing an example of the data type of credit principal overdue characteristic.

[0128] Further correlation analysis of feature data types under different categories showed that the correlation coefficients of all feature data types were less than 0.6. At the same time, when building the model, the correlation between feature data types would not have a significant impact on the model results. Fig.12 It is the correlation coefficient matrix of the credit principal overdue feature data type.

[0129] From the above analysis, it can be seen that there are great differences in the migration distribution patterns of account data under different guarantee types and overdue types. By extracting the risk characteristics of users' inability to repay account data in time through classification, the accuracy of user risk analysis can be further improved.

[0130] Then, the server can evaluate the first probability value corresponding to the account data based on the feature data. For example, the above-mentioned feature data of the user corresponding to the extracted account data sample can be used as input data, and the result of whether the corresponding account data sample is a bad sample or a good sample can be used as output. The model is trained to build a risk assessment model. Then, risk assessment can be performed based on the constructed risk assessment model. In some embodiments, the server can also construct risk identification models of different categories based on feature data of different categories, and then integrate the identification results of the risk identification models under each category to determine the risk of the user corresponding to the current account data. The specific implementation method is described in the following steps:

[0131] Inputting the credit principal overdue characteristic data into a credit principal risk identification model to obtain a first risk identification score;

[0132] Inputting the credit interest overdue characteristic data into a credit interest risk identification model to obtain a second risk identification score;

[0133] Inputting the guarantee overdue characteristic data into a guarantee risk identification model to obtain a third risk identification score;

[0134] A first probability value corresponding to the current account data is determined according to the first risk identification score, the second risk identification score, and the third risk identification score.

[0135] The server can use the sample data of the credit principal account data, the sample data of the credit interest account data and the sample data of the guarantee account data as input, and use the result of whether the corresponding account data sample is a good sample or a bad sample as output, and respectively construct the risk identification model of the corresponding category to obtain the credit principal risk identification model, the credit interest risk identification model and the guarantee risk identification model. The algorithm used in the model construction can be a binary classification model construction algorithm such as logistic regression.

[0136] Logistic regression is mainly used to solve binary classification problems in supervised learning. Its mathematical expression model is:

[0137]

[0138] Where θ is a parameter, the intuitive explanation of the output is: h θ (x) For a given input x, the probability that y = 1. Mathematical representation:

[0139] h θ (x) = P (y = 1 | x; θ).

[0140] The loss function for logistic regression is:

[0141]

[0142] The gradient descent method can be used to solve it. The algorithm flow is as follows:

[0143] (1) Initialize θ (random initialization)

[0144] (2) Iteration:

[0145] (3) If j(θ) can continue to decrease, return to (2); if it stops decreasing, output the result.

[0146] In other implementations, the risk identification model may also be constructed using algorithms such as xgboost, support vector machine, neural network, etc.

[0147] After the initial model construction is completed, evaluation tests such as model discrimination and stability can be performed to further adjust the model parameters based on the evaluation test results. For example, for a model built using a logistic regression algorithm, the KS value can be used as an evaluation indicator of the model's discrimination. Alternatively, the AUC value can also be used to evaluate the quality of the model.

[0148] The calculation method of KS value is:

[0149] (1) Calculate the number of good and bad accounts in each scoring interval.

[0150] (2) Calculate the ratio of the cumulative number of good accounts to the total number of good accounts (good%) and the ratio of the cumulative number of bad accounts to the total number of bad accounts (bad%) in each scoring interval.

[0151] (3) Calculate the absolute value of the difference between the cumulative bad account percentage and the cumulative good account percentage in each scoring interval (cumulative good% - cumulative bad%), and then take the maximum value of these absolute values ​​to obtain the KS value of this scoring card.

[0152] AUC (Area Under Curve) is the area under the ROC (receiver operating characteristic curve). ROC is a curve drawn based on a series of different binary classification methods (cutoff values ​​or decision thresholds), with the true positive rate (sensitivity) as the ordinate and the false positive rate (1-specificity) as the abscissa. AUC can be obtained by summing the areas of each part under the ROC curve.

[0153] Take the credit principal risk identification model as an example. Fig.13 , Fig.14 and Fig.15 As shown, Fig.13 A schematic diagram showing the discrimination of the credit principal risk identification model built based on training samples. Fig.14 A schematic diagram showing the discrimination of the risk identification model of the credit principal model after optimization based on the test sample. Fig.15 The diagram shows the discrimination of the credit principal risk identification model based on the validation sample. Among them, TPR represents the ratio of the real positive samples currently classified as positive samples to all positive samples, and FPR represents the ratio of the real negative samples currently incorrectly classified as positive samples to all negative samples. The KS value of the training sample and the test sample is 0.55, and the KS value of the validation set is 0.53, both of which have strong discrimination. The AUC of the training sample and the test sample exceeds 87%, and the AUC of the validation sample is 85.7%. The model PSI (Population Stability Index) is 0.0094, which is much less than 0.1, and the results are stable and reliable.

[0154] After the risk identification model and fusion processing model under each category are built, the stability monitoring, effectiveness monitoring and business monitoring of the overall model can also be carried out. Among them, the indicators for monitoring the stability of the model can be CSI (characteristic stability index) to measure the changes in the characteristic level, and PSI to measure the changes in the score distribution. The main indicators for monitoring the effectiveness of the model can be variable IV and model KS / AUC, which have the following functions: monitoring whether the risk identification ability of a single input variable is attenuated; evaluating whether the overall prediction ability of the model is reduced. In business monitoring, monitoring can be carried out from two aspects: bad debt rate and recovery rate. Bad debt rate monitoring is to monitor whether the bad debt rate corresponding to the risk value of different data segments is sortable. Recovery rate monitoring is to observe whether the recovery rate corresponding to the risk value of different data segments is sortable and whether it is consistent with expectations. The monitoring cycle can be set to one test per month. Through the above monitoring, the determination of risk value can be more in line with the dynamic development of actual business and the accuracy of risk prediction can be improved.

[0155] When predicting the risk of the user corresponding to the current account data, the business data of the user corresponding to the current account data and the data under the object label system can be respectively extracted. Then, the extracted credit principal overdue feature data, credit interest overdue feature data and guarantee overdue feature data can be respectively input into the credit principal risk identification model, the credit interest risk identification model and the guarantee risk identification model to obtain the first risk identification score, the second risk identification score and the third risk identification score. Afterwards, the first risk identification score, the second risk identification score and the third risk identification score can be fused, and the risk value after the fusion processing can be used to quantitatively evaluate the risk of the user. The fusion processing method can adopt algorithms such as decision trees and GBDT. The algorithm parameters corresponding to the fusion processing can also be optimized through training samples and test samples.

[0156] From the above analysis, we can see that there are great differences in the migration distribution of account data under different guarantee types and overdue types. By extracting feature data through classification and identifying the risk of users not being able to repay account data in time, the extracted feature data can more accurately represent the risk of users. At the same time, for account data, due to its own few data features, it is not very obvious to represent its risk of becoming a bad debt. Based on the corresponding user data features to assist in risk assessment, the accuracy of determining the risk of account data becoming a bad debt can be further improved.

[0157] It should be noted that the above scheme provided in the embodiments of this specification can be adapted to a variety of existing application scenarios. It is particularly suitable for small and micro enterprises. Financial institutions currently provide a variety of different loan products for such enterprises, which makes the loan data of such enterprises very different from the commonly used loan products, and the amount of data is also small. The collection methods of currently commonly used loan products may not be applicable to such enterprises. Based on the scheme provided in the above embodiments of this specification, the collection recovery rate of such enterprises and the reasonable allocation of collection resources can be greatly improved, while also improving the user experience.

[0158] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. For details, please refer to the description of the above-mentioned related processing related embodiments, and no further description is given here.

[0159] The above is a description of a specific embodiment of the specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0160] The method for screening the processing method of account data provided by one or more embodiments of this specification can analyze the account data from the perspective of guarantee method, overdue type, and the first probability value that the user corresponding to the overdue account cannot repay the account data in time, and divide the collection queues corresponding to each account data. Correspondingly, each collection queue has similar basic data characteristics and risk characteristics. Then, the second probability value of the account data that has not been settled during the corresponding overdue period can be quantified using the distribution of the bad debt ratio of each collection queue with the overdue time. Afterwards, the second probability value can be used to configure the processing method of the account data that has not been settled during the corresponding overdue period, thereby achieving differentiated configuration of the processing methods for account data with different risks, greatly improving the collection recovery rate of account data and the reasonable allocation of limited collection resources, improving the efficiency of account data processing, and reducing the risk of resource loss of financial institutions.

[0161] Based on the above-mentioned accounting data processing method screening method, one or more embodiments of this specification also provide an accounting data processing method screening device. The device may include a system, software (application), module, component, server, etc. that uses the method described in the embodiments of this specification and is combined with necessary implementation hardware. Based on the same innovative concept, the device in one or more embodiments provided in the embodiments of this specification is as described in the following embodiments. Since the implementation scheme of the device to solve the problem is similar to the method, the implementation of the specific device in the embodiments of this specification can refer to the implementation of the aforementioned method, and the repetitions will not be repeated. As used below, the term "unit" or "module" can be a combination of software and / or hardware that implements predetermined functions. Although the devices described in the following embodiments are preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceived. Specifically, Fig.16 The module structure diagram of an embodiment of an accounting data processing method screening device provided in the specification is shown as follows: Fig.16 As shown, the device may include:

[0162] The data receiving module 102 may be used to receive an account data processing method screening request, wherein the account data processing method screening request includes an account identifier; the account identifier is used to identify the current account data corresponding to the account data processing method screening request;

[0163] The data acquisition module 104 may be used to acquire a first probability value, a guarantee type, and an overdue type corresponding to the account identifier; the first probability value is used to represent the probability that the user cannot repay the amount of resources involved in the account data within a preset time;

[0164] The collection queue determination module 106 may be used to determine the collection queue corresponding to the current account data according to the first probability value, guarantee type and overdue type corresponding to the account identifier; the collection queue is pre-divided according to the first probability value, guarantee type and overdue type of the account data;

[0165] The probability value determination module 108 can be used to determine a second probability value that the current account data may become a bad debt at the corresponding overdue time by using the downward migration rate of the collection queue to which the current account data belongs at each overdue time; the downward migration rate is the ratio of the number of bad debts of the collection queue to the number of accounts of the collection queue at the corresponding overdue time; the bad debt includes account data that cannot be recovered within a preset time; the overdue time includes a time value exceeding the specified repayment time of the account data;

[0166] The processing method screening module 110 may be used to screen out the processing method corresponding to the current account data at the corresponding overdue time from the account data processing method set according to the second probability value.

[0167] In some other embodiments, the device may further include:

[0168] An acquisition module, which can be used to acquire the credit principal overdue characteristic data, credit interest overdue characteristic data and guarantee overdue characteristic data of the user corresponding to the current account data;

[0169] A first identification module may be used to input the credit principal overdue characteristic data into a credit principal risk identification model to obtain a first risk identification score;

[0170] The second identification module may be used to input the credit interest overdue characteristic data into a credit interest risk identification model to obtain a second risk identification score;

[0171] The third identification module can be used to input the overdue characteristic data of the guarantee into the guarantee risk identification model to obtain a third risk identification score;

[0172] The risk value determination module may be configured to determine a first probability value corresponding to the current account data according to the first risk identification score, the second risk identification score, and the third risk identification score.

[0173] In some other embodiments, the processing method screening module 110 may include:

[0174] The latest collection time acquisition unit can be used to obtain the latest collection time of the collection queue to which the current account data belongs under different processing modes; wherein the latest collection time is determined according to the distribution characteristics of the account data recovery rate of each collection queue relative to the overdue time;

[0175] The processing method screening unit can be used to determine the processing method adopted by the current account data at each overdue time according to the latest collection time and the second probability value.

[0176] In some other embodiments, the processing method may include a first collection channel, a second collection channel, and a third collection channel; wherein the probability of recovering the account data of the first collection channel, the second collection channel, and the third collection channel decreases in sequence.

[0177] In some other embodiments, the processing mode screening unit may include:

[0178] The first judgment subunit may be used to judge whether the current account data satisfies a first collection condition, wherein the first collection condition includes: a second probability value of the current account data becoming a bad debt at a specified overdue time is greater than or equal to a first collection threshold, or the specified overdue time is greater than or equal to a first latest collection time corresponding to the collection queue to which the current account data belongs under a first collection channel;

[0179] The first processing method screening subunit can be used to execute the first collection channel starting from the specified overdue time if yes.

[0180] In some other embodiments, the processing method screening unit may further include:

[0181] The second processing method screening sub-unit can be used to execute collection channels other than the first collection channel from the designated overdue time to N days after the designated overdue time if the designated overdue time is the first overdue day of the current account data; and execute the first collection channel starting from N days after the designated overdue time, where N is greater than or equal to 1.

[0182] In some other embodiments, the processing method screening unit may further include:

[0183] The second judgment subunit may be used to judge whether the current account data satisfies a second collection condition, wherein the second collection condition includes: a second probability value of the current account data becoming a bad debt at a specified overdue time is greater than or equal to a second collection threshold and less than a first collection threshold, or the specified overdue time is greater than or equal to a second latest collection time corresponding to the collection queue to which the current account data belongs under a second collection channel; wherein the first collection threshold is greater than the second collection threshold, and the first latest collection time is greater than the second latest collection time;

[0184] The third processing method screening subunit can be used to, if yes, execute the second collection channel from the specified overdue time to the first latest collection time.

[0185] In some other embodiments, the processing method screening unit may further include:

[0186] The fourth processing method screening subunit may be used to execute the third collection channel when the current account data satisfies neither the first collection condition nor the second collection condition.

[0187] In some other embodiments, the device may further include a first collection threshold determination module, and the first collection threshold determination module may include:

[0188] The first calculation unit may be used to calculate the difference between the executable upper limit of the first collection account corresponding to the first collection channel and the amount of the to-be-collected accounts covered by the latest collection time corresponding to the first collection channel, to obtain a first account number;

[0189] The first maximum downward migration rate determination unit may be used to obtain the maximum downward migration rate of each account data settled before the latest collection time corresponding to the first collection channel, wherein the maximum downward migration rate is equal to the downward migration rate of the collection queue where each settled account data is located on the corresponding settlement date;

[0190] The first collection threshold determination unit can be used to sort the account data settled before the latest collection time corresponding to the first collection channel in descending order according to the maximum downward migration rate, and use the maximum downward migration rate of the account data ranked first as the first collection threshold.

[0191] In some other embodiments, the device may further include a second collection threshold determination module, and the second collection threshold determination module may include:

[0192] The second calculation unit may be used to calculate the difference between the executable upper limit of the second collection account corresponding to the second collection channel and the amount of the to-be-collected accounts covered by the latest collection time corresponding to the second collection channel, to obtain the second account number;

[0193] The second maximum downward migration rate determination unit may be used to obtain the maximum downward migration rate of each account data settled before the latest collection time corresponding to the second collection channel; wherein the maximum downward migration rate of the account data is equal to the downward migration rate of the collection queue where the settled account data is located on the corresponding account data settlement date;

[0194] The second collection threshold determination unit can be used to sort the account data settled before the latest collection time corresponding to the second collection channel in descending order according to the maximum downward migration rate, and use the maximum downward migration rate of the account data ranked at the second account number as the second collection threshold.

[0195] It should be noted that the above-mentioned device may also include other implementations according to the description of the method embodiment. The specific implementation methods can refer to the description of the relevant method embodiment, and will not be described one by one here.

[0196] The account data processing method screening device provided in one or more embodiments of this specification can analyze the account data from the perspective of the guarantee method, overdue type, and the first probability value that the user corresponding to the overdue account cannot repay the account data in time, and divide the collection queues corresponding to each account data. Correspondingly, each collection queue has similar basic data characteristics and risk characteristics. Then, the second probability value of the account data that has not been settled during the corresponding overdue period can be quantified using the distribution of the bad debt ratio of each collection queue along with the overdue time. Afterwards, the second probability value can be used to configure the processing method of the account data that has not been settled during the corresponding overdue period, thereby achieving differentiated configuration of the processing methods for account data with different risks, greatly improving the collection recovery rate of account data and the reasonable allocation of limited collection resources, improving the efficiency of account data processing, and reducing the risk of resource loss of financial institutions.

[0197] The methods or devices described in the above embodiments provided in this specification can implement business logic through computer programs and record them on storage media, and the storage media can be read and executed by computers to achieve the effects of the solutions described in the embodiments of this specification. Therefore, this specification also provides an accounting data processing method screening device, including a processor and a memory storing processor executable instructions, and when the instructions are executed by the processor, the steps of the method described in any of the above embodiments are implemented.

[0198] The identity authentication device provided in the embodiments of this specification can be a computer terminal, a server or a similar computing device. Taking the server as an example, Fig.17 is a hardware structure block diagram of a server to which the embodiments of this specification are applied. Fig.17 As shown, the server 100 may include one or more (only one is shown in the figure) processors 200 (the processor 200 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA), a memory 300 for storing data, and a transmission module 400 for communication functions. A person skilled in the art can understand that Fig.17 The structure shown is only for illustration and does not limit the structure of the above electronic device. Fig.17 More or fewer components shown in the figure may also include other processing hardware, such as a database or multi-level cache, GPU, or other hardware with Fig.17 Show different configurations.

[0199] The memory 300 can be used to store software programs and modules of application software, such as program instructions / modules corresponding to the search method in the embodiment of the present invention. The processor 200 executes various functional applications and data processing by running the software programs and modules stored in the memory 300. The memory 300 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 300 may further include a memory remotely arranged relative to the processor 200, and these remote memories may be connected to the computer terminal via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0200] The transmission module 400 is used to receive or send data via a network. The specific example of the above network may include a wireless network provided by a communication provider of a computer terminal. In one example, the transmission module 400 includes a network adapter (Network Interface Controller, NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission module 400 can be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0201] The storage medium may include a physical device for storing information, which is usually a medium that digitizes the information and then stores it in an electrical, magnetic or optical manner. The storage medium may include: a device that stores information in an electrical energy manner, such as various memories, such as RAM, ROM, etc.; a device that stores information in a magnetic energy manner, such as a hard disk, a floppy disk, a magnetic tape, a magnetic core memory, a magnetic bubble memory, a USB flash drive; a device that stores information in an optical manner, such as a CD or a DVD. Of course, there are other readable storage media, such as quantum memory, graphene memory, etc.

[0202] It should be noted that the above-mentioned device may also include other implementations according to the description of the method embodiment. The specific implementation methods can refer to the description of the relevant method embodiment, and will not be described one by one here.

[0203] The account data processing method screening device described in the above embodiment can analyze the account data from the perspective of the guarantee method, overdue type, and the first probability value that the user corresponding to the overdue account cannot repay the account data in time, and divide the collection queues corresponding to each account data. Correspondingly, each collection queue has similar basic data characteristics and risk characteristics. Then, the second probability value of the account data that has not been settled during the corresponding overdue period can be quantified using the distribution of the bad debt ratio of each collection queue with the overdue time. Afterwards, the second probability value can be used to configure the processing method of the account data that has not been settled during the corresponding overdue period, thereby achieving differentiated configuration of the processing methods of account data with different risks, greatly improving the collection recovery rate of account data and the reasonable allocation of limited collection resources, improving the efficiency of account data processing, and reducing the risk of resource loss of financial institutions.

[0204] This specification also provides a system for screening accounting data processing methods, which can be a separate accounting data processing method screening system, and can also be applied to a variety of computer data processing systems. The system can be a separate server, or it can include a server cluster, system (including distributed system), software (application), actual operation device, logic gate circuit device, quantum computer, etc. that use one or more of the methods or one or more embodiments of this specification, and a terminal device combined with necessary implementation hardware. The accounting data processing method screening system can include a resource management system and a server, and the resource management system includes one or more business subsystems, wherein the resource management system can be used to send resource transfer data to the server; the server can include at least one processor and a memory storing computer executable instructions, and the processor implements the steps of the method described in one or more embodiments when executing the instructions; one or more business subsystems of the resource management system can be used to receive the identification results, and to manage and control the accounts of the currently monitored account type.

[0205] It should be noted that the above-mentioned system may also include other implementation methods according to the description of the method or device embodiment. The specific implementation method can refer to the description of the relevant method embodiment, which will not be described one by one here.

[0206] The account data processing method screening system described in the above embodiment can analyze the account data from the perspective of guarantee method, overdue type, and the first probability value that the user corresponding to the overdue account cannot repay the account data in time, and divide the collection queues corresponding to each account data. Correspondingly, each collection queue has similar basic data characteristics and risk characteristics. Then, the second probability value of the account data that has not been settled during the corresponding overdue period can be quantified using the distribution of the bad debt ratio of each collection queue with the overdue time. Afterwards, the second probability value can be used to configure the processing method of the account data that has not been settled during the corresponding overdue period, thereby achieving differentiated configuration of the processing methods for account data with different risks, greatly improving the collection recovery rate of account data and the reasonable allocation of limited collection resources, improving the efficiency of account data processing, and reducing the risk of resource loss of financial institutions.

[0207] The embodiments of this specification are not limited to situations that must comply with standard data models / templates or the embodiments of this specification. Certain industry standards or slightly modified implementation plans based on the implementation described in the custom methods or embodiments can also achieve the same, equivalent or similar, or predictable implementation effects after deformation of the above-mentioned embodiments. The embodiments obtained by using these modified or deformed data acquisition, storage, judgment, processing methods, etc. can still fall within the scope of the optional implementation plans of this specification.

[0208] Each embodiment in this specification is described in a progressive manner, and the same and similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment. In the description of this specification, the description of the reference terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of this specification. In this specification, the schematic representation of the above terms does not necessarily target the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner. In addition, in the absence of contradiction, a person skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples.

[0209] The above description is only an embodiment of the present specification and is not intended to limit the present specification. For those skilled in the art, the present specification may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present specification shall be included in the scope of the claims of the present specification.

Claims

1. A method for screening accounting data processing methods, characterized in that: Applicable to servers, including: Receiving an account data processing method screening request, wherein the account data processing method screening request includes an account identifier; the account identifier is used to identify current account data corresponding to the account data processing method screening request; Obtaining a first probability value, a guarantee type, and an overdue type corresponding to the account identifier; the first probability value is used to represent the probability that the user cannot repay the amount of resources involved in the account data within a preset time; Determine the collection queue corresponding to the current account data according to the first probability value, guarantee type and overdue type corresponding to the account identifier; the collection queue is pre-divided according to the first probability value, guarantee type and overdue type of the account data; The second probability value that the current account data may become a bad debt at the corresponding overdue time is determined by using the downward migration rate of the collection queue to which the current account data belongs at each overdue time; the downward migration rate is the ratio of the number of bad debts of the collection queue to the number of accounts of the collection queue at the corresponding overdue time; the bad debt includes account data that cannot be recovered within a preset time; the overdue time includes a time value exceeding the specified repayment time of the account data; Filtering out the processing method corresponding to the current account data at the corresponding overdue time from the account data processing method set according to the second probability value; The first probability value corresponding to the account identifier is determined in the following manner: Acquire the credit principal overdue characteristic data, credit interest overdue characteristic data and guarantee overdue characteristic data of the user corresponding to the account identifier; Inputting the credit principal overdue characteristic data into a credit principal risk identification model to obtain a first risk identification score; Inputting the credit interest overdue characteristic data into a credit interest risk identification model to obtain a second risk identification score; Inputting the guarantee overdue characteristic data into a guarantee risk identification model to obtain a third risk identification score; A first probability value corresponding to the account identifier is determined according to the first risk identification score, the second risk identification score, and the third risk identification score.

2. The method according to claim 1, characterized in that The step of selecting the processing method corresponding to the current account data at the corresponding overdue time from the account data processing method set according to the second probability value includes: Obtaining the latest collection time of the collection queue to which the current account data belongs under different processing modes; wherein the latest collection time is determined according to the distribution characteristics of the account data recovery rate of each collection queue relative to the overdue time; The processing method used for the current account data at each overdue time is determined according to the latest collection time and the second probability value.

3. The method according to claim 2, characterized in that The processing method includes a first collection channel, a second collection channel and a third collection channel; wherein the probability of recovering the account data of the first collection channel, the second collection channel and the third collection channel decreases in sequence.

4. The method according to claim 3, characterized in that The determining, according to the latest collection time and the second probability value, a processing method to be adopted for the current account data at each overdue time includes: Determining whether the current account data satisfies a first collection condition, wherein the first collection condition includes: a second probability value of the current account data becoming a bad debt at a specified overdue time is greater than or equal to a first collection threshold, or the specified overdue time is greater than or equal to a first latest collection time corresponding to the collection queue to which the current account data belongs under a first collection channel; If so, the first collection channel is executed starting from the specified overdue time.

5. The method according to claim 4, characterized in that The method further comprises: If the designated overdue time is the first overdue day of the current account data, collection channels other than the first collection channel are executed from the designated overdue time to N days after the designated overdue time; and the first collection channel is executed starting from N days after the designated overdue time, where N is greater than or equal to 1.

6. The method according to claim 4, characterized in that The determining, according to the latest collection time and the second probability value, a processing method to be adopted for the current account data at each overdue time includes: Determine whether the current account data satisfies a second collection condition, wherein the second collection condition includes: a second probability value of the current account data becoming a bad debt at a specified overdue time is greater than or equal to a second collection threshold and less than a first collection threshold, or the specified overdue time is greater than or equal to a second latest collection time corresponding to the collection queue to which the current account data belongs under a second collection channel; wherein the first collection threshold is greater than the second collection threshold, and the first latest collection time is greater than the second latest collection time; If so, the second collection channel is executed from the designated overdue time to the first latest collection time.

7. The method according to claim 6, characterized in that When the current account data satisfies neither the first collection condition nor the second collection condition, the third collection channel is executed.

8. The method according to claim 4, characterized in that The first collection threshold is determined in the following manner: Calculate the difference between the executable upper limit of the first collection account corresponding to the first collection channel and the amount of the to-be-collected accounts covered by the latest collection time corresponding to the first collection channel to obtain the first account number; Obtaining a maximum downward migration rate of each account data settled before the latest collection time corresponding to the first collection channel, wherein the maximum downward migration rate is equal to the downward migration rate of the collection queue where each settled account data is located on the corresponding settlement date; The account data settled before the latest collection time corresponding to the first collection channel are sorted in descending order according to the maximum downward migration rate, and the maximum downward migration rate of the account data ranked first is used as the first collection threshold.

9. The method according to claim 6, characterized in that The second collection threshold is determined in the following manner: Calculate the difference between the executable upper limit of the second collection account corresponding to the second collection channel and the amount of the to-be-collected accounts covered by the latest collection time corresponding to the second collection channel to obtain the second account number; Obtaining the maximum downward migration rate of each account data settled before the latest collection time corresponding to the second collection channel; wherein the maximum downward migration rate is equal to the downward migration rate of the collection queue where the settled account data is located on the settlement date of the corresponding account data; The account data settled before the latest collection time corresponding to the second collection channel are sorted in descending order according to the maximum downward migration rate, and the maximum downward migration rate of the account data ranked second is used as the second collection threshold.

10. A device for screening accounting data processing methods, characterized in that: Applicable to servers, including: A data receiving module, used for receiving a request for screening an accounting data processing method, wherein the request for screening an accounting data processing method includes an accounting identifier; the accounting identifier is used for identifying current accounting data corresponding to the request for screening an accounting data processing method; A data acquisition module, used to acquire a first probability value, a guarantee type, and an overdue type corresponding to the account identifier; the first probability value is used to represent the probability that the user cannot repay the amount of resources involved in the account data within a preset time; A collection queue determination module, used to determine the collection queue corresponding to the current account data according to the first probability value, guarantee type and overdue type corresponding to the account identifier; the collection queue is pre-divided according to the first probability value, guarantee type and overdue type of the account data; A probability value determination module is used to determine a second probability value that the current account data may become a bad debt at the corresponding overdue time by using the downward migration rate of the collection queue to which the current account data belongs at each overdue time; the downward migration rate is the ratio of the number of bad debts of the collection queue to the number of accounts of the collection queue at the corresponding overdue time; the bad debt includes account data that cannot be recovered within a preset time; the overdue time includes a time value that exceeds the specified repayment time of the account data; A processing method screening module is used to screen out the processing method corresponding to the current account data at the corresponding overdue time from the account data processing method set according to the second probability value.

11. A system for screening accounting data processing methods, characterized in that: The system comprises at least one processor and a memory for storing processor-executable instructions, wherein the instructions, when executed by the processor, implement the steps of the method according to any one of claims 1 to 9.

Images