Method, device, computer equipment and storage medium for determining object loss probability

Abstract

The present application relates to a method, device, computer equipment, storage medium and computer program product for determining the probability of object loss, and relates to the field of Internet financial technology. The method includes: obtaining each business identifier of the target object; determining whether the target business corresponding to each business identifier expires within the rated period; if each target business expires within the rated period, inputting the object information of the target object into a pre-trained first prediction model, and using the first prediction model to determine the first loss probability of the target object; if there is a business that has not expired within the rated period among each target business, inputting the object information of the target object into a pre-trained second prediction model, and using the second prediction model to determine the second loss probability of the target object. The use of this method can improve the accuracy of determining the probability of object loss, avoid misjudgment of the object loss situation, and improve the effectiveness of object maintenance.

Description

Technical Field

[0001] The present application relates to the field of Internet financial technology, and in particular to a method, apparatus, computer equipment, storage medium and computer program product for determining the probability of object loss. Background Art

[0002] At present, due to their short operating history and few mortgage assets, small and micro enterprises usually use short-term credit as their main source of debt financing. Unlike customers of medium- and long-term mortgage loans, small and micro loan customers generally need to adjust their short-term liabilities every year to adapt to their operating conditions, so their liquidity for short-term credit is strong. After a short-term loan of a small and micro loan customer expires, they can generally choose to sign a new short-term loan with the original financial institution (credit supplier) or go to other institutions to sign a new short-term loan. Therefore, for credit suppliers such as banks, small and micro loan customers are an important customer group that needs to be continuously maintained.

[0003] In the traditional technical solution, the prediction model is trained in advance using the sample object information of the sample object, and then after obtaining the object information of the target object, the object information is input into the prediction model, and the prediction model is used to predict the loss probability of the target object, so as to formulate a maintenance strategy based on the loss probability, perform object maintenance in advance, and prevent object loss. However, in actual operation, it is not accurate to directly use the prediction model to determine the loss probability for different objects, which will cause misjudgment of the object loss situation and affect the effectiveness of object maintenance.

[0004] Therefore, how to improve the accuracy of determining the probability of object loss, avoid misjudgment of the object loss situation, and improve the effectiveness of object maintenance is a technical problem that technical personnel in this field currently need to solve. Summary of the invention

[0005] Based on this, it is necessary to provide a method, device, computer equipment, computer-readable storage medium and computer program product for determining the probability of object loss, which can improve the accuracy of determining the probability of object loss, avoid misjudgment of object loss, and improve the effectiveness of object maintenance in response to the above-mentioned technical problems.

[0006] In a first aspect, the present application also provides a method for determining the probability of object loss. The method comprises:

[0007] Obtain each business identifier of the target object;

[0008] Determining whether the target services corresponding to the service identifiers are all expired within a rated period;

[0009] If all the target services expire within the rated period, the target object information is input into a pre-trained first prediction model, and the first prediction model is used to determine the first churn probability of the target object; the first prediction model is a model trained using the first sample object information of sample objects whose sample services all expire within the rated period;

[0010] If there is a business among the target businesses that has not expired within the rated period, the object information of the target object is input into a pre-trained second prediction model, and the second prediction model is used to determine the second churn probability of the target object; the second prediction model is a model trained using the second sample object information of the sample objects whose sample businesses have not expired within the rated period.

[0011] In one embodiment, the process of obtaining the first prediction model includes:

[0012] Acquire the first training sample; the first training sample includes the first sample object information of the sample object whose sample services are all due within the rated period and label information corresponding to the first sample object information;

[0013] The first training sample is input into an initial neural network for learning and training to obtain the first prediction model.

[0014] In one of the embodiments, the first sample object information includes: first sample characteristic information corresponding to information before the sample object handles the sample service, and second sample characteristic information corresponding to information after the sample object handles the sample service.

[0015] In one embodiment, the initial neural network includes a first initial neural network and a second initial neural network; inputting the first training sample into the initial neural network for learning and training to obtain the first prediction model includes:

[0016] Using the first sample feature information and corresponding label information, training the first initial neural network to obtain a corresponding first sub-model;

[0017] Using the second sample feature information and the corresponding label information, training the second initial neural network to obtain a corresponding second sub-model;

[0018] The first prediction model is obtained according to the first sub-model and the second sub-model.

[0019] In one of the embodiments, the first sample object information includes basic information of the sample object and association information corresponding to the sample object.

[0020] In one embodiment, the process of obtaining the first sample object information includes:

[0021] The obtained original data information is transparently transmitted, cross-compared or statistically aggregated to determine the first sample object information.

[0022] In one embodiment, before determining whether the target services respectively corresponding to the service identifiers have all expired within a rated period, the method further includes:

[0023] Determining whether the target object meets the hard churn condition;

[0024] If the target object meets the hard churn condition, determining the target object as an expired churn object;

[0025] Otherwise, return to the step of determining whether the target services respectively corresponding to the service identifiers have all expired within the rated period.

[0026] In a second aspect, the present application also provides a device for determining the probability of object loss. The device comprises:

[0027] An acquisition module, used to acquire each business identifier of a target object;

[0028] A determination module, used to determine whether the target services respectively corresponding to the service identifiers have all expired within a rated period;

[0029] A first execution module is used for inputting the object information of the target object into a pre-trained first prediction model if all the target services expire within the rated period, and determining a first churn probability of the target object using the first prediction model; the first prediction model is a model trained using first sample object information of sample objects whose sample services all expire within the rated period;

[0030] The second execution module is used to input the object information of the target object into a pre-trained second prediction model if there is a business that has not expired within the rated period among the target businesses, and use the second prediction model to determine the second churn probability of the target object; the second prediction model is a model trained using the second sample object information of the sample object whose sample business has not expired within the rated period.

[0031] In a third aspect, the present application further provides a computer device. The computer device includes a memory and a processor, the memory stores a computer program, and the processor implements the following steps when executing the computer program:

[0032] Obtain each business identifier of the target object;

[0033] Determining whether the target services corresponding to the service identifiers are all expired within a rated period;

[0034] If all the target services expire within the rated period, the target object information is input into a pre-trained first prediction model, and the first prediction model is used to determine the first churn probability of the target object; the first prediction model is a model trained using the first sample object information of sample objects whose sample services all expire within the rated period;

[0035] If there is a business among the target businesses that has not expired within the rated period, the object information of the target object is input into a pre-trained second prediction model, and the second prediction model is used to determine the second churn probability of the target object; the second prediction model is a model trained using the second sample object information of the sample objects whose sample businesses have not expired within the rated period.

[0036] In a fourth aspect, the present application further provides a computer-readable storage medium. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the following steps are implemented:

[0037] Obtain each business identifier of the target object;

[0038] Determining whether the target services corresponding to the service identifiers are all expired within a rated period;

[0039] If all the target services expire within the rated period, the target object information is input into a pre-trained first prediction model, and the first prediction model is used to determine the first churn probability of the target object; the first prediction model is a model trained using the first sample object information of sample objects whose sample services all expire within the rated period;

[0040] If there is a business among the target businesses that has not expired within the rated period, the object information of the target object is input into a pre-trained second prediction model, and the second prediction model is used to determine the second churn probability of the target object; the second prediction model is a model trained using the second sample object information of the sample objects whose sample businesses have not expired within the rated period.

[0041] In a fifth aspect, the present application further provides a computer program product. The computer program product includes a computer program, and when the computer program is executed by a processor, the following steps are implemented:

[0042] Obtain each business identifier of the target object;

[0043] Determining whether the target services corresponding to the service identifiers are all expired within a rated period;

[0044] If all the target services expire within the rated period, the target object information is input into a pre-trained first prediction model, and the first prediction model is used to determine the first churn probability of the target object; the first prediction model is a model trained using the first sample object information of sample objects whose sample services all expire within the rated period;

[0045] If there is a business among the target businesses that has not expired within the rated period, the object information of the target object is input into a pre-trained second prediction model, and the second prediction model is used to determine the second churn probability of the target object; the second prediction model is a model trained using the second sample object information of the sample objects whose sample businesses have not expired within the rated period.

[0046] The above-mentioned method, device, computer equipment, storage medium and computer program product for determining the probability of object churn, after obtaining each business identifier of the target object and determining whether the target business corresponding to each business identifier has expired within the rated period, if each target business has expired within the rated period, the object information of the target object is input into a pre-trained first prediction model, and the first prediction model is used to determine the first churn probability of the target object; if there is a business that has not expired within the rated period among the target businesses, the object information of the target object is input into a pre-trained second prediction model, and the second prediction model is used to determine the second churn probability of the target object; this method determines the corresponding second churn probability according to different situations of the target business of the target object, that is, the situation where all businesses expire or part of the businesses expire, respectively using the pre-trained first prediction model or the second prediction model. A churn probability or a second churn probability; since in this scheme the first prediction model is trained using the first sample object information of the sample objects whose sample businesses all expire within the rated period, and the second prediction model is trained using the second sample object information of the sample objects whose sample businesses do not expire within the rated period, the first prediction model and the second prediction model respectively correspond to the target business whose all businesses are due or the target business whose some businesses are due. Therefore, according to different situations of whether the target businesses of the target objects all expire within the rated period and whether there are businesses that have not expired within the rated period, it will be more accurate to determine the first churn probability or the second churn probability of the target object using the corresponding first prediction model or the second prediction model, which can improve the accuracy of the determined churn probability of the target object, avoid misjudgment of the object churn situation, and improve the effectiveness of object maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 A schematic diagram of a flow chart of a method for determining the probability of object loss in one embodiment;

[0048] Figure 2 A schematic diagram of a process for determining a first prediction model and a second prediction model in one embodiment;

[0049] Figure 3 A schematic diagram of a process for determining a first prediction model in one embodiment;

[0050] Figure 4 is a flow chart of a method for determining the probability of object loss in another embodiment;

[0051] Figure 5 is a structural block diagram of a device for determining the probability of object loss in one embodiment;

[0052] Figure 6 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0053] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0054] The method for determining the probability of object loss provided in the embodiment of the present application can be applied to a terminal or a server, or to a system including a terminal and a server. The terminal can be, but is not limited to, various personal computers, laptops, smart phones, tablet computers, and Internet of Things devices, and the server can be implemented as an independent server or a server cluster composed of multiple servers.

[0055] In one embodiment, Figure 1 As shown, a method for determining the probability of object loss is provided, which is described by taking the method applied to a terminal as an example, and includes the following steps:

[0056] Step 102: Obtain each service identifier of the target object.

[0057] Specifically, the target object refers to the object that has handled the target service. The target object can be an enterprise or an individual user. This embodiment does not limit the specific type of the target object. Among them, the service identifier refers to the identification information corresponding to the target service. Generally speaking, the unique corresponding service can be determined based on the service identifier. Specifically, the service identifier can be the name, identification code and other information of the service. This embodiment does not limit the specific type of the service identifier.

[0058] In actual operation, the object identifier of the target object is first obtained, and then the business identifier corresponding to the object identifier of the target object is determined according to the preset correspondence between the object identifier and the business identifier, that is, each business identifier of the target object is determined.

[0059] Step 104: determine whether the target services corresponding to the service identifiers are all expired within the rated period.

[0060] Specifically, according to each business identifier, determine the target business corresponding to each business identifier; then judge whether each corresponding target business expires within the rated period according to the business information of the target business. The expiration of the target business within the rated period means that the difference between the deadline of the time limit corresponding to the target business and the current time is within the rated period; for example, assuming that the time limit of a certain target business is 2 years (April 2020 to April 2022), that is, the deadline of the time limit of the target business is April 2022, that is, the target business will expire in April 2022. If the current time is March 2022, the difference between the deadline and the current time is 1 month. Assuming the rated period is 90 days, it means that the target business will expire within 90 days of the rated period.

[0061] After judging each target business separately and determining whether each target business will expire within the rated period, it is further determined whether all target businesses expire within the rated period. That is to say, only when all target businesses expire within the rated period, it means that all target businesses expire within the rated period; if any target business does not expire within the rated period, it means that there is a business among the target businesses that has not expired within the rated period.

[0062] Step 106, if all target businesses expire within the rated period, the object information of the target object is input into a pre-trained first prediction model, and the first prediction model is used to determine the first churn probability of the target object; the first prediction model is a model trained using the first sample object information of sample objects whose sample businesses all expire within the rated period.

[0063] Specifically, in this step, the first sample object information is used in advance for learning and training to obtain the corresponding first prediction model; the first sample object information is the object information corresponding to the sample objects whose sample services are all due within the rated period. If all target services are due within the rated period, the object information of the target object is input into the pre-trained first prediction model, and the first prediction model is used to output the corresponding first churn probability according to the object information of the target object, thereby obtaining the first churn probability of the target object.

[0064] Step 108, if there are businesses among the target businesses that have not expired within the rated period, the object information of the target object is input into a pre-trained second prediction model, and the second prediction model is used to determine the second churn probability of the target object; the second prediction model is a model trained using the second sample object information of sample objects whose sample businesses have not expired within the rated period.

[0065] Specifically, in this step, the second sample object information is used in advance for learning and training to obtain the corresponding second prediction model; the second sample object information is the object information corresponding to the sample object whose sample business does not expire within the rated period. If there is a business that does not expire within the rated period among the target businesses, the object information of the target object is input into the pre-trained second prediction model, and the second prediction model is used to output the corresponding second churn probability according to the object information of the target object, thereby obtaining the second churn probability of the target object.

[0066] The present embodiment provides a method for determining the probability of object churn. After obtaining each service identifier of the target object and determining whether the target services corresponding to each service identifier have all expired within the rated period, if each target service has expired within the rated period, the object information of the target object is input into a pre-trained first prediction model, and the first prediction model is used to determine the first churn probability of the target object; if there is a service that has not expired within the rated period among the target services, the object information of the target object is input into a pre-trained second prediction model, and the second prediction model is used to determine the second churn probability of the target object; according to different situations of the target services of the target object, that is, the situation where all services expire or part of the services expire, the present method uses the pre-trained first prediction model or the second prediction model to determine the corresponding first churn probability or the second churn probability. 2. Churn probability: Since in this scheme the first prediction model is trained using the first sample object information of the sample objects whose sample businesses all expire within the rated period, and the second prediction model is trained using the second sample object information of the sample objects whose sample businesses do not expire within the rated period, the first prediction model and the second prediction model respectively correspond to the target business whose all businesses are due or the target business whose part of the business is due. Therefore, according to different situations of whether the target businesses of the target objects all expire within the rated period and whether there are businesses that have not expired within the rated period, it will be more accurate to determine the first churn probability or the second churn probability of the target object using the corresponding first prediction model or the second prediction model, which can improve the accuracy of the determined churn probability of the target object, avoid misjudgment of the object churn situation, and improve the effectiveness of object maintenance.

[0067] Based on the above embodiment, this embodiment further illustrates and optimizes the technical solution. Specifically, in this embodiment, the process of obtaining the first prediction model includes:

[0068] Acquire a first training sample; the first training sample includes first sample object information of sample objects whose sample services are all due within a rated period and label information corresponding to the first sample object information;

[0069] The first training sample is input into the initial neural network for learning and training to obtain a first prediction model.

[0070] Among them, the first training sample refers to the sample used to train the first prediction model, and the first training sample includes first sample object information and label information corresponding to the first sample object information; the first sample object information refers to the object information corresponding to the sample object whose sample business expires within the rated period; the label information refers to the actual churn probability of the sample object corresponding to the first sample object information.

[0071] Specifically, in this embodiment, the initial neural network is predetermined, and the initial neural network may be a convolutional neural network (CNN), a recurrent neural network (RNN), a long short-term memory neural network (LSTM), a feed-forward neural network (FNNs), and a LightGBM scoring model, etc. This embodiment does not limit the specific type of the initial neural network.

[0072] In actual operation, the first training sample is input into the initial neural network for learning and training, and when the trained neural network reaches the expected accuracy, the trained neural network is determined as the first prediction model. In addition, the first sample object information in the first training sample can be feature information of the sample object, or can be information obtained after normalizing the feature information of the sample object.

[0073] It should be noted that the process of training the second prediction model is similar to the process of training the first prediction model, the difference is that the samples used for training are different; that is, the second training samples are input into the initial neural network for learning and training to obtain the second prediction model; the second training samples include second sample object information of sample objects whose sample business does not expire within the rated period and label information corresponding to the second sample object information; the process of training the second prediction model can refer to the above-mentioned process of training the first prediction model, and will not be repeated here.

[0074] The first prediction model or the second prediction model is trained and obtained in the manner of this embodiment, and the operation process is convenient and easy.

[0075] Based on the above embodiments, this embodiment further illustrates and optimizes the technical solution. Specifically, in this embodiment, the first sample object information includes: first sample characteristic information corresponding to the information before the sample object handles the sample business, and second sample characteristic information corresponding to the information after the sample object handles the sample business.

[0076] In this embodiment, the time node for sample objects whose sample services all expire within the rated period to handle sample services is used to obtain sample characteristic information of the sample objects before the time node, that is, to obtain the corresponding first sample characteristic information; and the sample characteristic information of the sample objects after the time node is obtained, that is, to obtain the corresponding second sample characteristic information.

[0077] It should be noted that the first sample characteristic information and the second sample characteristic information can be different data dimensions, that is, there is a new data dimension after the time node of handling the sample business, and the data information corresponding to the new data dimension is obtained; the first sample characteristic information and the second sample characteristic information can be the same data dimension, and the data information corresponding to the data dimension changes before and after the time node of handling the sample business.

[0078] For example, assuming that the sample business is a loan business, the first sample characteristic information corresponding to the information before the sample object handles the sample business includes the data information corresponding to the eight data dimensions of enterprise taxation, enterprise credit, enterprise capital transaction information, enterprise loan, enterprise owner credit card, enterprise owner personal credit, enterprise owner capital transaction, and enterprise owner personal loan before handling the loan business; the second sample characteristic information includes the data information corresponding to the eight data dimensions of enterprise taxation, enterprise credit, enterprise capital transaction information, enterprise loan, enterprise owner credit card, enterprise owner personal credit, enterprise owner capital transaction, and enterprise owner personal loan after handling the loan business.

[0079] In this embodiment, the first sample feature information corresponding to the information before the sample object handles the sample business, and the second sample feature information corresponding to the information after the sample object handles the sample business, are used to train a first prediction model or a second prediction model, so that the trained first prediction model can determine the churn probability of the target object by using more previous information.

[0080] It should be noted that the second sample object information includes: the third sample characteristic information corresponding to the information before the sample object handles the sample business, and the fourth sample characteristic information corresponding to the information after the sample object handles the sample business. Specifically, the sample object that has a business that has not expired within the rated period handles the sample business is used as the time node, and the sample characteristic information of the sample object before the time node is obtained, that is, the corresponding third sample characteristic information is obtained; and the sample characteristic information of the sample object after the time node is obtained, that is, the corresponding fourth sample characteristic information is obtained. The specific data dimension types of the third sample characteristic information and the fourth sample characteristic information are not limited here. For details, please refer to the above explanation of the first sample characteristic information and the second sample characteristic information, which will not be repeated here.

[0081] Based on the above embodiments, this embodiment further illustrates and optimizes the technical solution. Specifically, in this embodiment, the initial neural network includes a first initial neural network and a second initial neural network; the first training sample is input into the initial neural network for learning and training to obtain a first prediction model, including:

[0082] Using the first sample feature information and the corresponding label information, training the first initial neural network to obtain a corresponding first sub-model;

[0083] Using the second sample feature information and the corresponding label information, training the second initial neural network to obtain a corresponding second sub-model;

[0084] A first prediction model is obtained according to the first sub-model and the second sub-model.

[0085] In this embodiment, the first sample feature information and the second sample feature information are used to train the corresponding first sub-model and the second sub-model respectively. Specifically, the first sample feature information and the corresponding label information are used to train the first initial neural network to obtain the corresponding first sub-model; the second sample feature information and the corresponding label information are used to train the second initial neural network to obtain the corresponding second sub-model.

[0086] Among them, the first initial neural network and the second initial neural network can be the same neural network, such as Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN), Long Short-term Memory Networks (LSTM), Feed-forward Neural Networks (FNNs) and LightGBM scoring model, etc.

[0087] After determining the first sub-model and the second sub-model, a first prediction model is obtained according to the first sub-model and the second sub-model. Specifically, the output results of the first sub-model and the output results of the second sub-model are used for comprehensive calculation, and the result of the comprehensive calculation is used as the churn probability of the target object.

[0088] More specifically, in an actual operation, the first sample feature information and the second sample feature information are encoded respectively, and based on the LightGBM scoring model, the encoded data information is used for learning and training to obtain the corresponding first sub-model and second sub-model; the output results corresponding to the first sub-model and the second sub-model are encoded respectively, and input into the preset fusion probability model to obtain the corresponding first churn probability. The fusion probability model is established as follows:

[0089] y=σ(ω0+ω1X1+ω2X2);

[0090]

[0091] Among them, X1 represents the value after encoding the output result of the first sub-model, X2 represents the value after encoding the output result of the second sub-model; ω0 represents the value after encoding the model coefficient; ω1 and ω2 represent the values ​​after encoding the model coefficients of the first sub-model and the second sub-model respectively.

[0092] It can be seen that according to the method of this embodiment, the first prediction model is determined using the first sub-model and the second sub-model, so that the first prediction model can determine the first churn probability of the target object more comprehensively and accurately.

[0093] It can be understood that the second prediction model can also be determined according to similar processing steps, the difference is that the second sample object information used to train the second prediction model includes third sample feature information and fourth sample feature information; the third sample feature information and the corresponding label information are used to train the third initial neural network to obtain the corresponding third sub-model; the fourth sample feature information and the corresponding label information are used to train the fourth initial neural network to obtain the corresponding fourth sub-model; based on the third sub-model and the fourth sub-model, the second prediction model is obtained; this embodiment does not elaborate on the specific process of determining the second prediction model.

[0094] According to the method of this embodiment, the second prediction model is determined using the third sub-model and the fourth sub-model, so that the second prediction model can determine the second churn probability of the target object more comprehensively and accurately.

[0095] On the basis of the above embodiments, this embodiment further illustrates and optimizes the technical solution. Specifically, in this embodiment, the first sample object information includes basic information of the sample object and association information corresponding to the sample object.

[0096] Among them, basic information refers to the characteristics related to the characteristics of the sample object itself; for example, assuming that the sample object is an enterprise, the corresponding basic information includes the enterprise's own business conditions and credit history data; the associated information corresponding to the enterprise includes the personal information, credit behavior and other data of the actual controller (owner) of the enterprise. For another example, if the sample object is an individual, the corresponding basic information refers to the personal information, credit behavior and other data corresponding to the sample object, and the associated information with the individual includes the personal information, credit behavior and other data of the close relatives of the sample object.

[0097] In this embodiment, the first sample object information includes basic information of the sample object and associated information corresponding to the sample object, so the first prediction model obtained through training can determine the first churn probability of the target object by using more comprehensive information.

[0098] It should be noted that the second sample object information also includes the basic information of the sample object and the associated information corresponding to the sample object. The specific explanation of the second sample object information can refer to the above explanation of the first sample information, which will not be repeated here. Therefore, the second prediction model obtained by training can use more comprehensive information to determine the second churn probability of the target object.

[0099] Based on the above embodiment, this embodiment further illustrates and optimizes the technical solution. Specifically, in this embodiment, the process of obtaining the first sample object information includes:

[0100] The acquired original data information is transparently transmitted, cross-compared or statistically aggregated to determine the first sample object information.

[0101] Among them, the original data includes object information corresponding to sample objects whose sample services are all due within the rated period and object information of sample objects whose sample services are not due within the rated period; object information that can be used as training samples is obtained by transparent transmission, cross comparison or statistical aggregation operations on the original data; first sample object information and second sample object information are determined based on the processed object information and the corresponding label information.

[0102] Among them, pass-through is a processing method that directly uses data. When each sample object has only one data record, the data can be directly used or a simple data processing can be performed to generate a label. For data items of numerical type, the data can be directly used as a label. For example, the "People's Bank of China Credit Interpretation Score" feature of the business owner, the data source is "Personal Credit Report-Rating Information Unit-Credit Interpretation Score Data Item", this data item represents the personal credit score, and the score is directly used as a label as an indicator to measure the overall credit level of the reporting subject. For categorical data items, usually according to the specific meaning of each category, categories with similar meanings are merged as needed to generate labels. For example, the "education level" feature of the business owner, the data source is "Personal Credit Report-Identity Information Unit-Education Level Code Data Item". According to the code value mapping, seven categories are obtained: unknown, junior high school and below, technical secondary school, high school, junior college, undergraduate, and graduate school. According to the education level, technical secondary school, high school and high school are merged into one category, and junior college and undergraduate are merged into one category to generate the final education level label.

[0103] Among them, cross-comparison features are used to describe the relevance and consistency of information from different data sources. For example, the feature of "whether the spouse of the business owner is a shareholder" is produced as follows: the first step is to extract the data item of "personal credit report-personal basic information category-marriage information unit-spouse ID number" of the business owner, and then compare this data item with the "business credit report-business basic information category-registered capital and major investor information unit-investor ID number data item" of the enterprise to determine whether the spouse of the actual controller of the enterprise holds shares.

[0104] Among them, statistical aggregation features are labels generated by processing and integrating data item information through statistical aggregation functions. For data items that generate multiple records over time for a report subject, it is suitable to apply statistical aggregation methods to extract key information and make labels. For this type of feature, different time windows are usually divided and statistical variables in each window are calculated. For numerical data items, statistics such as sum, mean, minimum, and maximum can be calculated. For example, "Personal Credit Report-Credit Transaction Information Detail Classification-Loan Account Information Unit-Loan Account Balance Data Item" records the balance of all loan accounts of customers. Based on this data item, the sum of all credit balances of customers can obtain the "Enterprise Owner's Total Credit Balance" feature; for categorical variables, the number of occurrences of the main types and the types of occurrence types are calculated. For example, "Personal Credit Report-Information Summary-Query Record Summary Data Block-Query Record Summary Information Unit" records the People's Bank of China query reason code, and the code values ​​are "02-Credit Approval", "03-Credit Card Approval", "08-Guarantee Qualification Review", etc. Processing the query reason can obtain the "Enterprise Owner Credit Approval Query Number in the Past Month" feature.

[0105] It can be seen that according to the method of this embodiment, the first sample object information is determined through transparent transmission, cross comparison or statistical aggregation operations, the original data is extensive, and the operation process is convenient and easy.

[0106] Based on the above embodiment, this embodiment further illustrates and optimizes the technical solution. Specifically, in this embodiment, before determining whether the target services corresponding to the service identifiers are all expired within the rated period, the method further includes:

[0107] Determine whether the target object meets the hard churn conditions;

[0108] If the target object meets the hard churn condition, the target object is determined to be an expired churn object;

[0109] Otherwise, return to the step of determining whether the target services corresponding to the service identifiers are all expired within the rated period.

[0110] Specifically, the target object meets the hard loss condition when there is an abnormality in the target object's own information, which causes the target object to be unable to continue to handle other businesses after the target business expires. Therefore, the target object can be directly determined as an expired loss object. Among them, the hard loss conditions include the target object being a blacklisted object, the target object has been deregistered, there are overdue cases in the target object's historical business, the target object is a subject in the list of abnormal business operations of enterprises, and the target object has a bad credit record.

[0111] In actual operation, when there are multiple hard churn conditions, when the target object meets any hard churn condition, the target object is determined to be an expired churn object, that is, the churn probability is 100%; or, when the number of hard churn conditions met by the target object reaches a preset number threshold, the target object is determined to be an expired churn object, that is, the churn probability is 100%.

[0112] According to the method of this embodiment, before using the first prediction model or the second prediction model to determine the churn probability of the target object, it can be judged in advance whether the target object meets the hard churn condition, and the expired churn objects can be determined quickly and directly.

[0113] In order to enable people in the technical field to better understand the technical solutions in this application, the technical solutions in the embodiments of this application are described in detail below in combination with actual application scenarios. In the embodiments of this application, the target business is the loan business. The specific steps include two parts: model training environment and model production environment:

[0114] Part 1: Model training environment.

[0115] like Figure 2As shown, it is a schematic diagram of the process of determining the first prediction model and the second prediction model in this embodiment. According to the underlying basic database table (data source) owned by each bank or financial institution, sample objects whose sample businesses are all due within the rated period and sample objects whose sample businesses are not due within the rated period are screened. Specifically, small and micro enterprise customers whose loan businesses are due and settled are selected as sample objects within a certain time range. If a small and micro enterprise customer has multiple loan businesses due and settled during this period, they are regarded as multiple different sample objects according to the number of loans; according to whether the sample object has a loan business that has not expired within 90 days, the sample object is divided into sample objects whose sample businesses are all due within the rated period and sample objects whose sample businesses are not due within the rated period. Among them, the underlying basic database table generally contains but is not limited to all or part of the following content data: basic information of the enterprise, basic information of the business owner, industrial and commercial data, credit information (business credit and personal credit), transaction flow, enterprise borrowing and lending in this institution, asset information, business owner borrowing and lending in this institution, asset information, and information on the agency services of the enterprise in this institution (tax payment, salary payment, etc.). The loan business expiration date is selected as the observation point, and 90 days is the sample performance period. If the sample object has not signed any new credit products in this institution during the sample performance period, it is regarded as a bad sample (lost customer), and if there is at least one contract signed, it is a good sample (no lost customer).

[0116] According to the time relationship between the data and the handling of the target business, the features are divided into two categories: the first sample feature information corresponding to the information before handling the sample business (pre-loan features) and the second sample feature information corresponding to the information after handling the sample business (mid-loan features); among them, the pre-loan features are the data features of the target object when handling the loan business, and the observation time point is the time point of signing the contract. For the target object, this part of the features will not change with the time point of calling the model in the future; the mid-loan features are the data features of the target object after handling the loan business, and the observation time point is the model calling time point. The feature values ​​are changing, and the latest values ​​need to be calculated daily for model use.

[0117] After obtaining the pre-loan features and mid-loan features, based on the original data of each data source, the feature layer data table is formed through the following three feature processing methods: transparent transmission, cross comparison and statistical aggregation. Specifically, enterprise feature information can include industrial and commercial information, credit information, capital flow and tax information, etc.; among them, credit information includes account information, usage, repayment and overdue status, etc. Specifically, personal feature information includes credit card information, credit information and agency payment information, etc.; among them, credit card information includes usage, repayment and overdue status, etc., and credit information includes mortgage, car loan and mortgage credit, etc.

[0118] Models are respectively established for two types of sample objects, namely, sample objects whose sample businesses are all due within the rated period and sample objects whose sample businesses are not due within the rated period. The modeling architecture is the same, and a first prediction model and a second prediction model are obtained.

[0119] Take the sample object whose sample business all expires within the rated period as an example. Figure 3 The figure shows a schematic diagram of the process of determining the first prediction model. From the eight data dimensions of enterprise taxation, enterprise credit, enterprise capital transaction information, enterprise loan, business owner credit card, business owner personal credit, business owner capital transaction, and business owner personal loan, select the first sample feature information (pre-loan feature) corresponding to the information before the sample business and the second sample feature information (loan feature) corresponding to the information after the sample business; perform WOE (Weight of Evidence) encoding on the first sample feature information and the second sample feature information respectively, and based on the LightGBM scoring model, use the encoded data information for learning and training to obtain the corresponding first sub-model and second sub-model; encode the output results corresponding to the first sub-model and the second sub-model respectively, and input them into the pre-set fusion probability model to obtain the corresponding first loss probability. Among them, the fusion probability model is established as follows:

[0120] y=σ(ω0+ω1X1+ω2X2);

[0121]

[0122] Among them, X1 represents the value after encoding the output result of the first sub-model, X2 represents the value after encoding the output result of the second sub-model; ω0 represents the value after encoding the model coefficient; ω1 and ω2 represent the values ​​after encoding the model coefficients of the first sub-model and the second sub-model respectively.

[0123] For sample objects whose sample businesses do not expire within the rated period, the second prediction model is determined in the same manner.

[0124] The second part is the model production environment.

[0125] When there is a target object (small micro loan customer) whose loan business will expire within 90 days, the first prediction model or the second prediction model is called daily to determine the churn probability of the target object.

[0126] Combination Figure 4The flowchart of a method for determining the probability of object loss is shown. First, it is determined whether the target object meets the hard loss conditions. The hard loss conditions include that the target object is a blacklist object, the target object has been cancelled, the target object is included in the list of abnormal business operations of enterprises, the target object has overdue payments, and the target object has a bad credit record.

[0127] If the target object meets the hard churn conditions, the target object is determined to be an expired churn object, that is, the churn probability is 100%; otherwise, determine whether each target business of the target object expires within the rated period; if each target business expires within the rated period, input the object information of the target object into a pre-trained first prediction model, and use the first prediction model to determine the first churn probability of the target object; if there is a business among the target businesses that has not expired within the rated period, input the object information of the target object into a pre-trained second prediction model, and use the second prediction model to determine the second churn probability of the target object.

[0128] The present embodiment provides a method for determining the probability of object churn. After obtaining each service identifier of the target object and determining whether the target services corresponding to each service identifier have all expired within the rated period, if each target service has expired within the rated period, the object information of the target object is input into a pre-trained first prediction model, and the first prediction model is used to determine the first churn probability of the target object; if there is a service that has not expired within the rated period among the target services, the object information of the target object is input into a pre-trained second prediction model, and the second prediction model is used to determine the second churn probability of the target object; according to different situations of the target services of the target object, that is, the situation where all services expire or part of the services expire, the present method uses the pre-trained first prediction model or the second prediction model to determine the corresponding first churn probability or the second churn probability. 2. Churn probability: Since in this scheme the first prediction model is trained using the first sample object information of the sample objects whose sample businesses all expire within the rated period, and the second prediction model is trained using the second sample object information of the sample objects whose sample businesses do not expire within the rated period, the first prediction model and the second prediction model respectively correspond to the target business whose all businesses are due or the target business whose part of the business is due. Therefore, according to different situations of whether the target businesses of the target objects all expire within the rated period and whether there are businesses that have not expired within the rated period, it will be more accurate to determine the first churn probability or the second churn probability of the target object using the corresponding first prediction model or the second prediction model, which can improve the accuracy of the determined churn probability of the target object, avoid misjudgment of the object churn situation, and improve the effectiveness of object maintenance.

[0129] It should be understood that, although the steps in the flowcharts involved in the above embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps is not strictly limited in order, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above embodiments may include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0130] Based on the same inventive concept, the embodiment of the present application also provides a device for determining the probability of object loss for implementing the method for determining the probability of object loss involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in the embodiments of one or more device for determining the probability of object loss provided below can refer to the limitations of the method for determining the probability of object loss above, and will not be repeated here.

[0131] In one embodiment, Figure 5 As shown, a device for determining the probability of object loss is provided, comprising: an acquisition module 502, a determination module 504, a first execution module 506 and a second execution module 508, wherein:

[0132] An acquisition module 502 is used to acquire each service identifier of a target object;

[0133] A determination module 504 is used to determine whether the target services corresponding to the service identifiers are all expired within a rated period;

[0134] The first execution module 506 is used to input the object information of the target object into a pre-trained first prediction model if all target services expire within the rated period, and use the first prediction model to determine the first churn probability of the target object; the first prediction model is a model trained using the first sample object information of the sample objects whose sample services all expire within the rated period;

[0135] The second execution module 508 is used to input the object information of the target object into a pre-trained second prediction model if there is a business that has not expired within the rated period among the target businesses, and use the second prediction model to determine the second churn probability of the target object; the second prediction model is a model trained using the second sample object information of the sample object whose sample business has not expired within the rated period. The device for determining the object churn probability provided by the embodiment of the present application has the same beneficial effect as the method for determining the object churn probability described above.

[0136] In one embodiment, the first execution module includes:

[0137] The acquisition submodule is used to acquire a first training sample; the first training sample includes first sample object information of sample objects whose sample services are all due within a rated period and label information corresponding to the first sample object information;

[0138] The input submodule is used to input the first training sample into the initial neural network for learning and training to obtain a first prediction model.

[0139] In one embodiment, the initial neural network includes a first initial neural network and a second initial neural network; the input submodule includes:

[0140] A first training unit is used to train the first initial neural network using the first sample feature information and the corresponding label information to obtain a corresponding first sub-model;

[0141] A second training unit is used to train a second initial neural network using the second sample feature information and the corresponding label information to obtain a corresponding second sub-model;

[0142] The determination unit is used to obtain a first prediction model according to the first sub-model and the second sub-model.

[0143] In one embodiment, the first training unit includes:

[0144] The acquisition subunit is used to perform transparent transmission, cross comparison or statistical aggregation operations on the acquired original data information to determine the first sample object information.

[0145] In one embodiment, the device further comprises:

[0146] A condition determination module is used to determine whether the target object meets the hard churn condition;

[0147] The conditional execution module is used to determine that the target object is an expired churn object if the target object meets the hard churn condition; otherwise, return to the determination module.

[0148] Each module in the above-mentioned object loss probability determination device can be implemented in whole or in part by software, hardware, or a combination thereof. Each of the above-mentioned modules can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute the operations corresponding to each of the above modules.

[0149] In one embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as follows: Figure 6As shown. The computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, a mobile cellular network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, a method for determining the probability of object loss is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covered on the display screen, or a button, trackball or touchpad set on the computer device housing, or an external keyboard, touchpad or mouse, etc.

[0150] Those skilled in the art will understand that Figure 6 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0151] In one embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the following steps are implemented:

[0152] Obtain each business identifier of the target object;

[0153] Determine whether the target services corresponding to the service identifiers are all expired within the specified period;

[0154] If all target businesses expire within the rated period, the target object's object information is input into a pre-trained first prediction model, and the first prediction model is used to determine the first churn probability of the target object; the first prediction model is a model trained using the first sample object information of sample objects whose sample businesses all expire within the rated period;

[0155] If there are businesses among the target businesses that have not expired within the rated period, the object information of the target object is input into a pre-trained second prediction model, and the second prediction model is used to determine the second churn probability of the target object; the second prediction model is a model trained using the second sample object information of sample objects whose sample businesses have not expired within the rated period.

[0156] A computer device provided in an embodiment of the present application has the same beneficial effects as the above-mentioned method for determining the probability of object loss.

[0157] In one embodiment, a computer readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:

[0158] Obtain each business identifier of the target object;

[0159] Determine whether the target services corresponding to the service identifiers are all expired within the specified period;

[0160] If all target businesses expire within the rated period, the target object's object information is input into a pre-trained first prediction model, and the first prediction model is used to determine the first churn probability of the target object; the first prediction model is a model trained using the first sample object information of sample objects whose sample businesses all expire within the rated period;

[0161] If there are businesses among the target businesses that have not expired within the rated period, the object information of the target object is input into a pre-trained second prediction model, and the second prediction model is used to determine the second churn probability of the target object; the second prediction model is a model trained using the second sample object information of sample objects whose sample businesses have not expired within the rated period.

[0162] A computer-readable storage medium provided in an embodiment of the present application has the same beneficial effects as the above-mentioned method for determining the probability of object loss.

[0163] In one embodiment, a computer program product is provided, comprising a computer program, which, when executed by a processor, implements the following steps:

[0164] Obtain each business identifier of the target object;

[0165] Determine whether the target services corresponding to the service identifiers are all expired within the specified period;

[0166] If all target businesses expire within the rated period, the target object's object information is input into a pre-trained first prediction model, and the first prediction model is used to determine the first churn probability of the target object; the first prediction model is a model trained using the first sample object information of sample objects whose sample businesses all expire within the rated period;

[0167] If there are businesses among the target businesses that have not expired within the rated period, the object information of the target object is input into a pre-trained second prediction model, and the second prediction model is used to determine the second churn probability of the target object; the second prediction model is a model trained using the second sample object information of sample objects whose sample businesses have not expired within the rated period.

[0168] A computer program product provided in an embodiment of the present application has the same beneficial effects as the above-mentioned method for determining the probability of object loss.

[0169] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0170] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.

[0171] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0172] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A method for determining the probability of object loss, characterized in that: The method comprises: Obtain each business identifier of the target object; Determining whether the target services corresponding to the service identifiers are all expired within a rated period; If all the target services expire within the rated period, the target object information is input into a pre-trained first prediction model, and the first prediction model is used to determine the first churn probability of the target object; the first prediction model is a model trained using the first sample object information of sample objects whose sample services all expire within the rated period; If there is a business that has not expired within the rated period among the target businesses, the object information of the target object is input into a pre-trained second prediction model, and the second prediction model is used to determine the second churn probability of the target object; the second prediction model is a model trained by using the second sample object information of the sample object whose sample business has not expired within the rated period; If the sample subject has not signed any new credit product with the institution during the sample performance period, the sample subject is a lost customer; if the sample subject has signed at least one contract during the sample performance period, the sample subject is a non-lost customer; The process of obtaining the first prediction model includes: Acquire a first training sample; the first training sample includes the first sample object information of the sample object whose sample services all expire within the rated period and label information corresponding to the first sample object information; the first sample object information includes: first sample feature information corresponding to information before the sample object handles the sample service, and second sample feature information corresponding to information after the sample object handles the sample service, wherein the second sample feature information has the same data dimension as the first sample feature information; Using the first sample feature information and the corresponding label information, training a first initial neural network to obtain a corresponding first sub-model; Using the second sample feature information and the corresponding label information, training a second initial neural network to obtain a corresponding second sub-model; The first prediction model is obtained according to the first sub-model and the second sub-model.

2. The method according to claim 1, characterized in that: The first sample object information includes basic information of the sample object and association information corresponding to the sample object.

3. The method according to claim 1, characterized in that: The process of obtaining the first sample object information includes: The obtained original data information is transparently transmitted, cross-compared or statistically aggregated to determine the first sample object information.

4. The method according to any one of claims 1 to 3, characterized in that: Before determining whether the target services respectively corresponding to the service identifiers have all expired within a rated period, the method further includes: Determining whether the target object meets the hard churn condition; If the target object meets the hard churn condition, determining the target object as an expired churn object; Otherwise, return to the step of determining whether the target services respectively corresponding to the service identifiers have all expired within the rated period.

5. A device for determining the probability of object loss, characterized in that: The device comprises: An acquisition module, used to acquire each business identifier of a target object; A determination module, used to determine whether the target services respectively corresponding to the service identifiers have all expired within a rated period; A first execution module is used for inputting the object information of the target object into a pre-trained first prediction model if all the target services expire within the rated period, and determining a first churn probability of the target object using the first prediction model; the first prediction model is a model trained using first sample object information of sample objects whose sample services all expire within the rated period; A second execution module is used for inputting the object information of the target object into a pre-trained second prediction model if there is a business that has not expired within the rated period among the target businesses, and determining the second churn probability of the target object by using the second prediction model; the second prediction model is a model trained by using the second sample object information of the sample object whose sample business has not expired within the rated period; If the sample subject has not signed any new credit product with the institution during the sample performance period, the sample subject is a lost customer; if the sample subject has signed at least one contract during the sample performance period, the sample subject is a non-lost customer; The acquisition submodule is used to acquire a first training sample; the first training sample includes the first sample object information of the sample object whose sample services all expire within the rated period and label information corresponding to the first sample object information; the first sample object information includes: first sample feature information corresponding to information before the sample object handles the sample service, and second sample feature information corresponding to information after the sample object handles the sample service, wherein the second sample feature information has the same data dimension as the first sample feature information; A first training unit, configured to train a first initial neural network using the first sample feature information and corresponding label information to obtain a corresponding first sub-model; A second training unit is used to train a second initial neural network using the second sample feature information and the corresponding label information to obtain a corresponding second sub-model; A determination unit is used to obtain the first prediction model according to the first sub-model and the second sub-model.

6. The device according to claim 5, characterized in that The first training unit comprises: The acquisition subunit is used to perform transparent transmission, cross comparison or statistical aggregation operations on the acquired original data information to determine the first sample object information.

7. The device according to claim 5 or 6, characterized in that Also includes: A condition determination module, used to determine whether the target object meets the hard churn condition; The conditional execution module is used to determine that the target object is an expired churn object if the target object meets the hard churn condition; otherwise, return to the determination module.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 4 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 4 are implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 4 are implemented.

Images