Overdue stage prediction method and device, electronic equipment and storage medium
By adopting a two-way LSTM network model in the financial collection scenario, predicting the overdue stage of users, solving the problems of large data processing volume and low prediction efficiency in the existing technology, and achieving more efficient and accurate prediction.
Patent Information
- Application Number
- CN202510217478.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-02-26
- Publication Date
- 2025-05-30
AI Technical Summary
The prior art requires processing a large amount of data when predicting whether a user will repay overdue, resulting in low prediction efficiency and difficulty in ensuring accuracy.
A two-way LSTM network model is used to obtain the overdue characteristics of the user during each historical overdue stage and input them into the pre-trained model to make predictions. This method reduces data processing volume, improves prediction efficiency, and captures users' overdue characteristics from both directions through a two-way network, enhancing prediction accuracy.
By reducing data processing volume, prediction efficiency is improved, and prediction accuracy is enhanced through the two-way LSTM network, which can more accurately predict the overdue phase of users.
Smart Images
Figure CN120070038A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of computer technology, and more particularly, to a method, device, electronic device, and storage medium for predicting overdue stages. Background Art
[0002] If a user fails to repay the loan within the specified period after taking the loan, it can be called overdue. For users in the overdue state, corresponding collection efforts will be made to urge them to repay the loan as soon as possible. Therefore, in the financial collection scenario, it is very important to predict whether a user will repay the loan overdue.
[0003] In the prior art, usually, data of users every day is collected to predict the overdue situation of users. This method requires processing a large amount of data, not only with low prediction efficiency but also difficult to ensure prediction accuracy. Summary of the Invention
[0004] In view of this, an object of the present invention is to provide a method, device, electronic device, and storage medium for predicting overdue stages, so as to improve prediction efficiency and enhance prediction accuracy.
[0005] To achieve the above object, the technical solutions adopted in the embodiments of the present invention are as follows:
[0006] In a first aspect, the present invention provides a method for predicting overdue stages, the method including:
[0007] Obtaining overdue characteristics within each historical overdue stage of a user; the historical overdue stage includes a plurality of consecutive days;
[0008] Inputting the overdue characteristics within each historical overdue stage into a pre-trained overdue stage prediction model; the overdue stage prediction model uses a bidirectional LSTM network;
[0009] Predicting the overdue stage at which the user will be overdue next time through the overdue stage prediction model.
[0010] In an optional implementation manner, the inputting the overdue characteristics within each historical overdue stage into a pre-trained overdue stage prediction model includes:
[0011] Inputting the overdue characteristics within each historical overdue stage into the forward LSTM network in the overdue stage prediction model according to a first order, and inputting the overdue characteristics within each historical overdue stage into the reverse LSTM network in the overdue stage prediction model according to a second order; the first order and the second order are determined according to the overdue sequence corresponding to each historical overdue stage, and the first order and the second order are opposite.
[0012] In an optional implementation manner, the overdue stage prediction model is trained in the following manner:
[0013] Generate a training data set based on the overdue characteristics within multiple historical overdue stages of different users;
[0014] Train an overdue stage prediction model based on a bidirectional LSTM network using the training data set to obtain a trained overdue stage prediction model.
[0015] In an alternative embodiment, the method further includes:
[0016] After the training of the overdue stage prediction model is completed, use the LRP algorithm to evaluate the importance of each overdue characteristic.
[0017] In an alternative embodiment, the use of the LRP algorithm to evaluate the importance of each overdue characteristic includes:
[0018] Based on the correlation of the current layer of the overdue stage prediction model, calculate the correlation calculation results from the current layer to the previous layer of the forward LSTM network in the overdue stage prediction model and the correlation calculation results from the current layer to the previous layer of the reverse LSTM network in the overdue stage prediction model respectively;
[0019] Perform a weighted calculation on the correlation calculation results from the current layer to the previous layer of the forward LSTM network and the correlation calculation results from the current layer to the previous layer of the reverse LSTM network to obtain the correlation of the previous layer of the current layer of the overdue stage prediction model;
[0020] Through layer-by-layer iterative calculation, finally obtain the correlation of each overdue characteristic of the input layer of the overdue stage prediction model; the correlation of each overdue characteristic of the input layer represents the importance of each overdue characteristic.
[0021] In an alternative embodiment, the calculation formula for the correlation calculation result from the current layer to the previous layer of the forward LSTM network is:
[0022]
[0023] where i represents the previous layer, j represents the current layer, R i←j represents the correlation calculation result from the current layer to the previous layer of the forward LSTM network, R j represents the correlation of the current layer of the overdue stage prediction model, z i represents the output of the i-th layer of the forward LSTM network, w ij represents the weight of the connection between the i-th and j-th layers of the forward LSTM network;
[0024] The calculation formula for the correlation calculation result from the current layer to the previous layer of the reverse LSTM network is:
[0025]
[0026] Among them, R' i←j represents the result of correlation calculation from the current layer to the previous layer of the reverse LSTM network, and z' i represents the output of the i-th layer of the reverse LSTM network, and w' ij represents the weight of the connection between the i-th and j-th layers of the reverse LSTM network.
[0027] In a second aspect, the present invention provides an overdue stage prediction device, and the device includes:
[0028] An acquisition module, configured to acquire overdue characteristics within each historical overdue stage of a user; the historical overdue stage includes a plurality of consecutive days;
[0029] An input module, configured to input the overdue characteristics within each historical overdue stage into a pre-trained overdue stage prediction model; the overdue stage prediction model uses a bidirectional LSTM network;
[0030] A prediction module, configured to predict the overdue stage where the user's next overdue will occur through the overdue stage prediction model.
[0031] In an optional implementation manner, the input module is configured to input the overdue characteristics within each historical overdue stage into the forward LSTM network in the overdue stage prediction model according to a first order, and input the overdue characteristics within each historical overdue stage into the reverse LSTM network in the overdue stage prediction model according to a second order; the first order and the second order are determined according to the overdue sequence corresponding to each historical overdue stage, and the first order and the second order are opposite.
[0032] In a third aspect, the present invention provides an electronic device, including a processor, a memory, and a computer program stored on the memory and executable on the processor, and when the computer program is executed by the processor, the steps of the overdue stage prediction method as described in any one of the foregoing embodiments are implemented.
[0033] In a fourth aspect, the present invention provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the overdue stage prediction method as described in any one of the foregoing embodiments are implemented.
[0034] The overdue stage prediction method, device, electronic device, and storage medium provided by the embodiments of the present invention, the method includes: obtaining the overdue characteristics within each historical overdue stage of the user; the historical overdue stage includes a plurality of consecutive days; inputting the overdue characteristics within each historical overdue stage into a pre-trained overdue stage prediction model; the overdue stage prediction model uses a bidirectional LSTM network; predicting the overdue stage where the user will be overdue next time through the overdue stage prediction model. By statistically analyzing the characteristics according to the overdue stage, compared with collecting the user's data every day to predict the user's overdue situation, the data processing volume is greatly reduced, and the prediction efficiency is improved; and by using a bidirectional LSTM network, the overdue characteristics of the user can be comprehensively captured from two directions, enhancing the prediction accuracy.
[0035] To make the above objects, features, and advantages of the present invention more obvious and understandable, the following specifically enumerates preferred embodiments and, in conjunction with the accompanying drawings, makes a detailed description as follows. BRIEF DESCRIPTION OF THE DRAWINGS
[0036] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required in the embodiments. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as limiting the scope. For those of ordinary skill in the art, other related drawings can be obtained based on these drawings without creative efforts.
[0037] Figure 1 Shows a schematic flowchart of an overdue stage prediction method provided by an embodiment of the present invention;
[0038] Figure 2 Shows another schematic flowchart of an overdue stage prediction method provided by an embodiment of the present invention;
[0039] Figure 3 Shows a functional module diagram of an overdue stage prediction device provided by an embodiment of the present invention;
[0040] Figure 4 Shows another functional module diagram of an overdue stage prediction device provided by an embodiment of the present invention;
[0041] Figure 5 Shows a schematic block diagram of an electronic device provided by an embodiment of the present invention.
[0042] Icons: 100 - electronic device; 110 - memory; 120 - processor; 130 - communication module; 600 - overdue stage prediction device; 610 - acquisition module; 620 - input module; 630 - prediction module; 640 - calculation module. DETAILED DESCRIPTION OF THE EMBODIMENTS
[0043] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. The components of the embodiments of the present invention described and illustrated herein can be arranged and designed in various different configurations.
[0044] Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative efforts belong to the scope of protection of the present invention.
[0045] It should be noted that relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprising", "including" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "comprising one..." does not exclude the existence of additional identical elements in the process, method, article or device comprising the element.
[0046] When traditional machine learning algorithms or time series models are used for overdue prediction, data of each day is usually used to predict the overdue situation of users. The amount of data to be processed is large, resulting in not only low prediction efficiency but also difficulty in ensuring prediction accuracy. In addition, the interpretability of the model is poor. For example, it is difficult for existing models to clearly explain why a certain user is overdue, making it difficult to provide an accurate interpretability basis for subsequent optimization of collection strategies.
[0047] Based on this, the embodiments of the present invention provide a method, device, electronic device and storage medium for overdue stage prediction. By statistically characterizing according to the overdue stage, compared with collecting the user's data every day to predict the user's overdue situation, the data processing volume is greatly reduced and the prediction efficiency is improved. And a bidirectional LSTM network is adopted, which can comprehensively capture the user's overdue characteristics from two directions, enhancing the prediction accuracy. In addition, by introducing the LRP (Layer-wise Relevance Propagation) algorithm to evaluate the importance of each overdue characteristic in the input layer of the model, it can be clearly known which overdue characteristics have the greatest impact on the prediction result, thereby improving the interpretability of the model and providing a more accurate interpretability basis for the subsequent optimization of the collection strategy.
[0048] Next, each embodiment of the present invention will be described in detail with reference to the accompanying drawings.
[0049] Please refer to Figure 1 , which is a schematic flowchart of a method for overdue stage prediction provided by an embodiment of the present invention. It should be noted that the method for overdue stage prediction of the present invention is not limited to Figure 1 the following specific order. It should be understood that in other embodiments, the order of some steps of the method for overdue stage prediction of the present invention can be interchanged according to actual needs, or some of the steps can also be omitted or deleted. The method for overdue stage prediction can be applied to electronic devices such as laptops, tablets, PCs (Personal Computers), and servers. Next, the Figure 1 specific process shown will be described in detail.
[0050] Step S101, obtain the overdue characteristics of the user in each historical overdue stage; the historical overdue stage includes multiple consecutive days.
[0051] In this embodiment, the overdue stage can be in units of months or quarters, and can be specifically set according to the actual business scenario. For example, each historical overdue stage of the user can include overdue for one month, two months, etc., which are respectively recorded as m1, m2.
[0052] For each historical overdue stage of the user, respectively count the overdue characteristics of the user in this historical overdue stage. The overdue characteristics can include: overdue days, overdue amount, amount to be repaid, click statistics of the loan App used, speed of the car gps, distance statistics, etc.
[0053] Step S102, input the overdue characteristics in each historical overdue stage into a pre-trained overdue stage prediction model; the overdue stage prediction model adopts a bidirectional LSTM network.
[0054] In this embodiment, the LSTM (Long Short-Term Memory) network is a special neural network structure that is good at processing time series data. In this embodiment, a bidirectional LSTM network is adopted, which can consider the information in both the forward time and the reverse time simultaneously, so as to more comprehensively understand the user's behavior pattern and improve the prediction accuracy of the future overdue stage.
[0055] Step S103, predict the overdue stage where the user will next be overdue through the overdue stage prediction model.
[0056] In this embodiment, the overdue stage prediction model adopts a bidirectional LSTM network. After inputting the overdue features in each historical overdue stage into the overdue stage prediction model, the overdue stage prediction model can comprehensively capture the user's overdue features from both the positive and negative directions, thereby enhancing the prediction accuracy.
[0057] It can be seen that the overdue stage prediction method provided by the embodiment of the present invention, by counting features according to the overdue stage, greatly reduces the data processing volume and improves the prediction efficiency compared with collecting the user's data every day to predict the user's overdue situation; and by adopting a bidirectional LSTM network, it can comprehensively capture the user's overdue features from two directions and enhance the prediction accuracy. In addition, by adopting phased features, it can not only provide an understandable basis for the explanation of subsequent collection strategies, but also simplify the time cost of understanding the importance of collection features.
[0058] In one implementation, the bidirectional LSTM network adopted by the overdue stage prediction model includes a forward LSTM network and a reverse LSTM network. Specifically, inputting the overdue features in each historical overdue stage into the pre-trained overdue stage prediction model in step S102 may include:
[0059] Input the overdue features in each historical overdue stage into the forward LSTM network in the overdue stage prediction model according to the first order, and input the overdue features in each historical overdue stage into the reverse LSTM network in the overdue stage prediction model according to the second order; the first order and the second order are determined according to the overdue sequence corresponding to each historical overdue stage, and the first order and the second order are opposite.
[0060] In this embodiment, the first order can be the order from the earliest overdue record to the most recent overdue record, and the second order can be the order traced back from the most recent overdue record. For example, a certain user has had 3 overdue experiences in the past. The overdue times of the 3 overdue records from the earliest to the most recent are m2 (assumed to be 2 months), m1 (assumed to be 1 month), and m3 (assumed to be 3 months). After obtaining the overdue characteristics of these three overdue stages of m2, m1, and m3, the corresponding overdue characteristics will be input into the forward LSTM network in the order of m2, m1, m3, and the corresponding overdue characteristics will be input into the reverse LSTM network in the order of m3, m1, m2.
[0061] It can be seen that in the embodiment of the present invention, a bidirectional LSTM network is adopted. By combining forward and reverse time series information, the model can learn more features, and thus make more accurate predictions, improving the prediction accuracy of the user's future overdue stages.
[0062] In one implementation, the overdue stage prediction model can be trained in the following manner:
[0063] Generate a training data set based on the overdue characteristics within multiple historical overdue stages of different users; train the overdue stage prediction model based on the bidirectional LSTM network according to the training data set to obtain a trained overdue stage prediction model.
[0064] Suppose a user has 4 overdue records in history. It is necessary to count the overdue characteristics of the four overdue stages and input them into the model for training. The overdue stages corresponding to the four overdue records are m2, m1, m3, and m1 respectively. Taking the overdue characteristics of the first overdue stage as an example, the following characteristics within the first overdue stage are counted: overdue days, overdue amount, amount to be repaid, statistics of clicks on the loan App used, speed of the car gps, distance statistics, etc. Generate training data and label each training data (i.e., the true overdue stage of the user's next overdue), thereby constructing a training data set.
[0065] When performing model training, taking the training data including the overdue characteristics of the four overdue stages of m2, m1, m3, and m1 as an example, the corresponding overdue characteristics can be input into the forward LSTM network in the order of m2, m1, m3, and m1, and the corresponding overdue characteristics can be input into the reverse LSTM network in the order of m1, m3, m1, and m2 to train the overdue stage prediction model until the overdue stage prediction model reaches the expected effect.
[0066] It can be seen that in the embodiment of the present invention, when training the model, the characteristics are statistically analyzed in terms of time by overdue stage, reducing the data explosion caused by counting characteristics by day, shortening the time required for model training, and saving a large amount of device resources.
[0067] In one embodiment, to help business personnel better understand the decision-making process of the model and ensure high interpretability of the model, the embodiments of the present invention introduce the LRP algorithm to evaluate the importance of each overdue feature in the input layer of the model, so as to clearly know which overdue features have the greatest impact on the prediction result, thereby improving the interpretability of the model and providing a more accurate interpretability basis for the subsequent optimization of the collection strategy.
[0068] Please refer to Figure 2 , the overdue stage prediction method provided by the embodiments of the present invention may further include:
[0069] Step S201, after the overdue stage prediction model is trained, use the LRP algorithm to evaluate the importance of each overdue feature.
[0070] In this embodiment, the LRP algorithm is mainly used for model interpretation and feature importance evaluation, that is, it is used to understand the decision-making process of the model after the model prediction is completed, so as to determine which input features have the greatest impact on the final prediction result. For example, in the historical overdue records of a certain user, there are features such as overdue days, overdue amount, and App usage habits. When using the trained overdue stage prediction model to predict the overdue stage of the user's next overdue, based on the LRP algorithm, it can be known which features (such as overdue days or App usage frequency) have the greatest impact on the prediction result.
[0071] In one embodiment, in the above step S201, using the LRP algorithm to evaluate the importance of each overdue feature may specifically include:
[0072] Based on the relevance of the current layer of the overdue stage prediction model, calculate the relevance calculation results from the current layer to the previous layer of the forward LSTM network in the overdue stage prediction model and the relevance calculation results from the current layer to the previous layer of the reverse LSTM network in the overdue stage prediction model respectively; perform weighted calculation on the relevance calculation results from the current layer to the previous layer of the forward LSTM network and the relevance calculation results from the current layer to the previous layer of the reverse LSTM network to obtain the relevance of the previous layer of the current layer of the overdue stage prediction model; through layer-by-layer iterative calculation, finally obtain the relevance of each overdue feature in the input layer of the overdue stage prediction model; the relevance of each overdue feature in the input layer represents the importance of each overdue feature.
[0073] In this embodiment, when using the LRP algorithm to evaluate the importance of overdue features, it starts from the output layer of the overdue stage prediction model and performs layer-by-layer iterative calculation to finally obtain the relevance of each overdue feature in the input layer of the overdue stage prediction model. Among them, the relevance of the output layer of the overdue stage prediction model can be set in advance.
[0074] In this embodiment, the calculation formula for the correlation calculation result from the current layer to the previous layer of the forward LSTM network is as follows:
[0075]
[0076] where i represents the previous layer, j represents the current layer, R i←j represents the correlation calculation result from the current layer to the previous layer of the forward LSTM network, R j represents the correlation of the current layer of the overdue stage prediction model, z i represents the output of the i-th layer of the forward LSTM network, w ij represents the weight of the connection between the i-th and j-th layers of the forward LSTM network.
[0077] The calculation formula for the correlation calculation result from the current layer to the previous layer of the reverse LSTM network is as follows:
[0078]
[0079] where R′ i←j represents the correlation calculation result from the current layer to the previous layer of the reverse LSTM network, z′ i represents the output of the i-th layer of the reverse LSTM network, w′ ij represents the weight of the connection between the i-th and j-th layers of the reverse LSTM network.
[0080] In this embodiment, after calculating the correlation calculation result R i←j from the current layer to the previous layer of the forward LSTM network in the overdue stage prediction model, and the correlation calculation result R′ i←j from the current layer to the previous layer of the reverse LSTM network in the overdue stage prediction model, through weighted calculation, the correlation R i of the previous layer of the current layer of the overdue stage prediction model is finally obtained, which can be specifically expressed as R i = A * R i←j + B * R′ i←j , where A and B are preset weight values. For example, A = 0.5 and B = 0.5 can be set.
[0081] After obtaining the correlation R i of the previous layer of the current layer of the overdue stage prediction model, through iterative calculation, the importance of each overdue feature in the input layer is obtained, so as to determine the importance of each overdue feature to the current prediction result.
[0082] It can be seen that the embodiment of the present invention proposes an interpretation method of bidirectional LRP for the overdue stage prediction model based on the bidirectional LSTM network, which can reduce the problem of forgetting the importance of features at the front during the interpretation of the unidirectional time series, provide an accurate interpretability basis for subsequent collection, and reduce the labor cost of updating strategies.
[0083] To execute the corresponding steps in the above embodiments and various possible ways, an implementation manner of an overdue stage prediction device is given below. Please refer to Figure 3 , which is a functional module diagram of an overdue stage prediction device 600 provided by an embodiment of the present invention. It should be noted that the basic principle and the technical effects generated by the overdue stage prediction device 600 provided in this embodiment are the same as those in the above embodiments. For a brief description, for the parts not mentioned in this embodiment, reference may be made to the corresponding content in the above embodiments. The overdue stage prediction device 600 includes: an acquisition module 610, an input module 620, and a prediction module 630.
[0084] The acquisition module 610 is used to acquire the overdue features of a user within each historical overdue stage; the historical overdue stage includes multiple consecutive days.
[0085] It can be understood that the acquisition module 610 can execute the above step S101.
[0086] The input module 620 is used to input the overdue features within each historical overdue stage into a pre-trained overdue stage prediction model; the overdue stage prediction model uses a bidirectional LSTM network.
[0087] It can be understood that the input module 620 can execute the above step S102.
[0088] The prediction module 630 is used to predict the overdue stage where the user will be overdue next time through the overdue stage prediction model.
[0089] It can be understood that the prediction module 630 can execute the above step S103.
[0090] Optionally, the input module 620 is used to input the overdue features within each historical overdue stage into the forward LSTM network in the overdue stage prediction model according to the first order, and input the overdue features within each historical overdue stage into the reverse LSTM network in the overdue stage prediction model according to the second order; the first order and the second order are determined according to the overdue sequence corresponding to each historical overdue stage, and the first order and the second order are opposite.
[0091] Optionally, the overdue stage prediction model is trained in the following manner:
[0092] Generate a training data set according to the overdue features within multiple historical overdue stages of different users;
[0093] Train an overdue stage prediction model based on a bidirectional LSTM network using a training dataset to obtain a trained overdue stage prediction model.
[0094] Optionally, refer to Figure 4 The overdue stage prediction device 600 may further include a calculation module 640, which is configured to evaluate the importance of each overdue feature using the LRP algorithm after the training of the overdue stage prediction model is completed.
[0095] It can be understood that the calculation module 640 may execute the above step S201.
[0096] Optionally, the calculation module 640 is specifically configured to calculate the correlation calculation result from the current layer to the previous layer of the forward LSTM network in the overdue stage prediction model and the correlation calculation result from the current layer to the previous layer of the backward LSTM network in the overdue stage prediction model respectively based on the correlation of the current layer of the overdue stage prediction model; perform a weighted calculation on the correlation calculation result from the current layer to the previous layer of the forward LSTM network and the correlation calculation result from the current layer to the previous layer of the backward LSTM network to obtain the correlation of the previous layer of the current layer of the overdue stage prediction model; and finally obtain the correlation of each overdue feature of the input layer of the overdue stage prediction model through layer-by-layer iterative calculation; the correlation of each overdue feature of the input layer represents the importance of each overdue feature.
[0097] Optionally, the formula for calculating the correlation calculation result from the current layer to the previous layer of the forward LSTM network is:
[0098]
[0099] where i represents the previous layer, j represents the current layer, R i←j represents the correlation calculation result from the current layer to the previous layer of the forward LSTM network, R j represents the correlation of the current layer of the overdue stage prediction model, z i represents the output of the i-th layer of the forward LSTM network, w ij represents the weight of the connection between the i-th and j-th layers of the forward LSTM network;
[0100] The formula for calculating the correlation calculation result from the current layer to the previous layer of the backward LSTM network is:
[0101]
[0102] where R′ i←j represents the correlation calculation result from the current layer to the previous layer of the backward LSTM network, z′ i represents the output of the i-th layer of the backward LSTM network, w′ij Represents the weight of the connection between the i-th and j-th layers of the reverse LSTM network.
[0103] It can be seen that the overdue stage prediction device provided by the embodiment of the present invention includes an acquisition module, an input module, and a prediction module. The acquisition module is used to acquire the overdue characteristics of the user in each historical overdue stage; the historical overdue stage includes multiple consecutive days; the input module is used to input the overdue characteristics in each historical overdue stage into a pre-trained overdue stage prediction model; the overdue stage prediction model adopts a bidirectional LSTM network; the prediction module is used to predict the overdue stage where the user will be overdue next time through the overdue stage prediction model. By statistically analyzing the characteristics according to the overdue stage, compared with collecting the user's data every day to predict the user's overdue situation, the amount of data processing is greatly reduced, and the prediction efficiency is improved; and by adopting a bidirectional LSTM network, the overdue characteristics of the user can be comprehensively captured from two directions, enhancing the prediction accuracy.
[0104] Please refer to Figure 5 , which is a schematic block diagram of an electronic device 100 provided by the embodiment of the present invention. The electronic device 100 includes a memory 110, a processor 120, and a communication module 130. The elements of the memory 110, the processor 120, and the communication module 130 are directly or indirectly electrically connected to each other to realize data transmission or interaction. For example, these elements can be electrically connected to each other through one or more communication buses or signal lines.
[0105] Among them, the memory 110 is used to store programs or data. The memory 110 can be, but is not limited to, a random access memory (RAM), a read only memory (ROM), a programmable read only memory (PROM), an erasable programmable read only memory (EPROM), an electrically erasable programmable read only memory (EEPROM), etc.
[0106] The processor 120 is used to read / write the data or programs stored in the memory 110 and execute corresponding functions. For example, when the computer program stored in the memory 110 is executed by the processor 120, the overdue stage prediction methods disclosed in the above embodiments can be realized.
[0107] The communication module 130 is used to establish a communication connection between the electronic device 100 and other devices through a network, and is used to send and receive data through the network.
[0108] It should be understood that Figure 5 the structure shown is only a schematic diagram of the structure of the electronic device 100, and the electronic device 100 may also include more or fewer components than Figure 5 shown therein, or have a configuration different from Figure 5 that shown. Figure 5 Each component shown therein may be implemented by hardware, software, or a combination thereof.
[0109] The embodiments of the present invention also provide a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by the processor 120, the overdue stage prediction method disclosed in the above embodiments is implemented.
[0110] In several embodiments provided in the present application, it should be understood that the disclosed apparatus and method may also be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the flowcharts and block diagrams in the drawings show the possible architectures, functions, and operations of apparatuses, methods, and computer program products according to multiple embodiments of the present invention. In this regard, each block in the flowchart or block diagram may represent a module, a program segment, or a part of code, and a module, a program segment, or a part of code contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in the reverse order, depending on the functions involved. It should also be noted that each block in the block diagram and / or flowchart, and the combination of blocks in the block diagram and / or flowchart, may be implemented by a dedicated hardware-based system for performing the specified functions or actions, or may be implemented by a combination of dedicated hardware and computer instructions.
[0111] In addition, in each embodiment of the present invention, the functional modules may be integrated together to form an independent part, or each module may exist separately, or two or more modules may be integrated to form an independent part.
[0112] If a function is implemented in the form of a software functional module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or a part of this technical solution can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods of various embodiments of the present invention. The aforementioned storage medium includes: various media that can store program codes such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs.
[0113] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.
Claims
1. A method for predicting the overdue stage, characterized in that: The method comprises: Obtaining overdue characteristics of the user in each historical overdue stage; the historical overdue stage includes multiple consecutive days; Inputting the overdue features in each of the historical overdue stages into a pre-trained overdue stage prediction model; the overdue stage prediction model adopts a bidirectional LSTM network; The overdue stage prediction model is used to predict the overdue stage that the user will be in next time he / she is overdue.
2. The overdue stage prediction method according to claim 1, characterized in that: The step of inputting the overdue features in each of the historical overdue stages into a pre-trained overdue stage prediction model comprises: The overdue features in each of the historical overdue stages are input into the forward LSTM network in the overdue stage prediction model in a first order, and the overdue features in each of the historical overdue stages are input into the reverse LSTM network in the overdue stage prediction model in a second order; the first order and the second order are determined according to the order of overdue corresponding to each of the historical overdue stages, and the first order is opposite to the second order.
3. The overdue stage prediction method according to claim 1, characterized in that: The overdue stage prediction model is trained in the following way: Generate a training data set based on the overdue features of different users in multiple historical overdue stages; The overdue stage prediction model based on the bidirectional LSTM network is trained according to the training data set to obtain a trained overdue stage prediction model.
4. The overdue stage prediction method according to any one of claims 1 to 3, characterized in that: The method further comprises: After the overdue stage prediction model training is completed, the LRP algorithm is used to evaluate the importance of each overdue feature.
5. The overdue stage prediction method according to claim 4, characterized in that: The LRP algorithm is used to evaluate the importance of each overdue feature, including: Based on the correlation of the current layer of the overdue stage prediction model, respectively calculate the correlation calculation results from the current layer to the previous layer of the forward LSTM network in the overdue stage prediction model, and the correlation calculation results from the current layer to the previous layer of the reverse LSTM network in the overdue stage prediction model; Performing weighted calculation on the correlation calculation result from the current layer to the previous layer of the forward LSTM network and the correlation calculation result from the current layer to the previous layer of the reverse LSTM network to obtain the correlation of the previous layer of the current layer of the overdue stage prediction model; Through iterative calculation layer by layer, the correlation of each overdue feature of the input layer of the overdue stage prediction model is finally obtained; the correlation of each overdue feature of the input layer represents the importance of each overdue feature.
6. The overdue stage prediction method according to claim 5, characterized in that: The calculation formula for the correlation calculation result from the current layer to the previous layer of the forward LSTM network is: Among them, i represents the previous layer, j represents the current layer, and R i←j Represents the correlation calculation result from the current layer to the previous layer of the forward LSTM network, R j represents the relevance of the current layer of the overdue stage prediction model, z i represents the output of the i-th layer of the forward LSTM network, w ij Represents the weight of the connection between the i-th and j-th layers of the forward LSTM network; The calculation formula for the correlation calculation result from the current layer to the previous layer of the reverse LSTM network is: Among them, R i ′ ←j represents the correlation calculation result from the current layer to the previous layer of the reverse LSTM network, z i ′ represents the output of the i-th layer of the reverse LSTM network, w i ′ j Represents the weight of the connection between the i-th and j-th layers of the reverse LSTM network.
7. A device for predicting overdue stage, characterized in that: The device comprises: An acquisition module, used to acquire overdue characteristics of the user in each historical overdue stage; the historical overdue stage includes a plurality of consecutive days; An input module, used for inputting the overdue features in each of the historical overdue stages into a pre-trained overdue stage prediction model; the overdue stage prediction model adopts a bidirectional LSTM network; The prediction module is used to predict the overdue stage of the user's next overdue payment through the overdue stage prediction model.
8. The overdue stage prediction device according to claim 7, characterized in that: The input module is used to input the overdue features in each of the historical overdue stages into the forward LSTM network in the overdue stage prediction model in a first order, and to input the overdue features in each of the historical overdue stages into the reverse LSTM network in the overdue stage prediction model in a second order; the first order and the second order are determined according to the overdue order corresponding to each of the historical overdue stages, and the first order is opposite to the second order.
9. An electronic device, characterized in that: The method comprises a processor, a memory and a computer program stored in the memory and executable on the processor, wherein when the computer program is executed by the processor, the steps of the overdue stage prediction method according to any one of claims 1 to 6 are implemented.
10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the overdue stage prediction method according to any one of claims 1 to 6 are implemented.