Credit data processing method and device, electronic equipment and storage medium

Abstract

The invention provides a credit data processing method and device, electronic equipment and a storage medium. The method comprises the steps of obtaining object credit data of a target object; inputting the object credit data into the trained long-short term sharing layer for feature extraction to obtain target long-term features and target short-term features; inputting the target short-term feature into a trained short-term prediction network to obtain a first overdue probability output by the short-term prediction network; inputting the target long-term feature into a trained long-term prediction network to obtain a second overdue probability output by the long-term prediction network; and determining the credit overdue risk of the target object based on the first overdue probability and / or the second overdue probability. The method can meet overdue prediction requirements of different risk observation periods.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, and in particular to a credit data processing method and device, electronic equipment and storage medium. BACKGROUND

[0002] In the credit risk control scene, how to accurately predict the credit overdue risk of a user is the key to credit risk control. Considering stability, the model side hopes to model with long-term risk (mob6+) labels; considering timeliness, the strategy side hopes to adjust based on short-term risk (such as mob3); how to balance the model performance requirements and strategy requirements is a problem that needs to be solved for credit risk control. SUMMARY

[0003] The present application provides a credit data processing method and device, electronic equipment and storage medium, which can distinguish long-term and short-term features for modeling and meet long-term and short-term risk prediction. The technical solution is as follows: According to one aspect of the present application, a credit data processing method is provided, the method comprising: obtaining object credit data of a target object; inputting the object credit data into a trained long-short term shared layer for feature extraction to obtain target long-term features and target short-term features; inputting the target short-term features into a trained short-term prediction network to obtain a first overdue probability output by the short-term prediction network; inputting the target long-term features into a trained long-term prediction network to obtain a second overdue probability output by the long-term prediction network; determining a credit overdue risk of the target object based on the first overdue probability and / or the second overdue probability.

[0004] According to another aspect of the present application, a credit data processing device is provided, the device comprising: a first obtaining module configured to obtain object credit data of a target object; a first feature extraction module configured to input the object credit data into a trained long-short term shared layer for feature extraction to obtain target long-term features and target short-term features; a first prediction module configured to input the target short-term features into a trained short-term prediction network to obtain a first overdue probability output by the short-term prediction network; a second prediction module configured to input the target long-term features into a trained long-term prediction network to obtain a second overdue probability output by the long-term prediction network; a determination module configured to determine a credit overdue risk of the target object based on the first overdue probability and / or the second overdue probability.

[0005] According to an aspect of the present application, an electronic device is provided, comprising a processor and a memory storing a program, the program comprising instructions which, when executed by the processor, cause the processor to perform the credit data processing method as described above.

[0006] According to another aspect of the present application, a non-transitory computer-readable storage medium storing computer instructions is provided, the computer instructions being used to cause a computer to perform the credit data processing method as described above.

[0007] According to another aspect of the present application, a computer program product is provided, the computer program product comprising computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions to cause the computer device to perform the credit data processing method as described above.

[0008] The technical solutions provided by the embodiments of the present application have at least the following beneficial effects: When it is necessary to predict the overdue risk of a target object, the object credit data of the target object can be obtained, and the object credit data is input into the trained long-short shared layer for feature extraction to obtain target long-term features and target short-term features. Then, the target short-term features are input into the trained short-term prediction network to obtain the first overdue probability output by the short-term prediction network; and the target long-term features are input into the trained long-term prediction network to obtain the second overdue probability output by the long-term prediction network. Then, the credit overdue risk of the target object is determined according to the first overdue probability and / or the second overdue probability. Through multi-task joint prediction, the long-short term overdue risk prediction of the target object can be realized, which can meet the overdue prediction demand of different risk observation periods on the strategy side while meeting the stability demand of model prediction. BRIEF DESCRIPTION OF DRAWINGS

[0009] In the following description of exemplary embodiments in conjunction with the accompanying drawings, more details, features and advantages of the present application are disclosed, in which: Figure 1 A flowchart of a credit data processing method according to an exemplary embodiment of the present application is shown; Figure 2 is a flowchart of a model training method provided by an exemplary embodiment of the present application; Figure 3 is a structural schematic diagram of a credit data processing device provided by an embodiment of the present application; Figure 4 A structural block diagram of an exemplary electronic device capable of implementing the embodiments of the present application is shown. DETAILED DESCRIPTION

[0010] Embodiments of the present application will be described in more detail with reference to the drawings. While certain embodiments of the present application are shown in the drawings, it is understood that the present application can be carried out in various forms and should not be construed as limited to the embodiments set forth herein, but rather should be construed to cover all modifications, equivalents, and alternatives falling within the spirit and scope of the present application. It is understood that the drawings and embodiments are only for illustrative purposes and not intended to limit the scope of the present application.

[0011] It should be understood that each step described in the method embodiments of the present application can be performed in different orders and / or in parallel. In addition, the method embodiments can include additional steps and / or omit the steps shown. The scope of the present application is not limited in this respect.

[0012] The term "comprising" and variations thereof as used in the present application are open-ended, that is, "including but not limited to". The term "based on" is "based, at least in part, on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". Related definitions are given throughout the description. It should be noted that the concepts mentioned in the present application as "first", "second", etc. are only used to distinguish different devices, modules or units, and are not intended to limit the order or interdependence of the functions performed by these devices, modules or units. It should be noted that the terms "one", "multiple" mentioned in the present application are illustrative and not limiting, and those skilled in the art should understand that "one" or "multiple" should be understood as "one or more" unless the context clearly indicates otherwise. The names of the messages or information exchanged between the devices in the embodiments of the present application are only for illustrative purposes and are not intended to limit the scope of the messages or information.

[0013] The scheme of the present application is described below with reference to the drawings, and the technical solutions provided by the embodiments of the present application are described in detail through specific embodiments and application scenarios.

[0014] Reference is made to Figure 1 which shows a flowchart of a credit data processing method according to an exemplary embodiment of the present application. Taking the case where the method is applied to an electronic device for example, as shown in Figure 1 , the method comprises: Step 101, obtaining object credit data of a target object; Step 102, inputting the object credit data into a trained long-short shared layer for feature extraction to obtain target long-term features and target short-term features; Step 103, input the target short-term feature into the trained short-term prediction network to obtain a first overdue probability output by the short-term prediction network; Step 104, input the target long-term feature into the trained long-term prediction network to obtain a second overdue probability output by the long-term prediction network; Step 105, determine the credit overdue risk of the target object based on the first overdue probability and / or the second overdue probability. To balance the model side and the strategy side, the present application independently learns short-term features and long-term features through multi-task learning to respectively perform short-term risk prediction and long-term risk prediction. In one possible implementation, a joint risk prediction model including a long-short-term shared layer + a short-term prediction network + a long-term prediction network is pre-trained, wherein the long-short-term shared layer is used to extract short-term features affecting short-term risk and long-term features affecting long-term risk, the weight of each feature in the short-term features is not the same as the weight of each feature in the long-term features, the weight of the feature affecting the short-term risk in the short-term features is greater than that of other features, and the weight of the feature affecting the long-term risk in the long-term features is greater than that of other features. The short-term prediction network is used to predict the overdue risk performance of the user at MOB3 (3 months after the loan), and the long-term prediction network is used to predict the overdue risk performance of the user at MOB6 (6 months after the loan).

[0015] After the joint risk prediction model is trained, the joint risk prediction model can be deployed to realize overdue risk prediction for different risk observation periods. In one possible implementation, when it is necessary to predict the overdue risk of a target object, object credit data of the target object can be obtained, which can include education information, income information, historical transaction information, historical credit information, historical debt information, age information, etc. of the target object; and the object credit data is input into the trained long-short-term shared layer for feature extraction to obtain target long-term features and target short-term features, the target short-term features are input into the trained short-term prediction network to obtain a first overdue probability output by the short-term prediction network, the first overdue probability represents the overdue risk performance of the target object at MOB3; and the target long-term features are input into the trained long-term prediction network to obtain a second overdue probability output by the long-term prediction network, the second overdue probability represents the overdue risk performance of the target object at MOB6, and the higher the first overdue probability, the higher the overdue risk of the target object during MOB3, the lower the first overdue probability, the lower the overdue risk of the target object during MOB3, the higher the second overdue probability, the higher the overdue risk of the target object during MOB6, and the lower the second overdue probability, the lower the overdue risk of the target object during MOB6.

[0016] After the overdue probability of the target object at MOB3 and MOB6 is predicted, the first overdue probability and the second overdue probability can be fed back to the business personnel, and the business personnel determines the credit overdue risk of the target object according to the first overdue probability and the second overdue probability. Alternatively, a first probability weight can be set for MOB3 and a second probability weight can be set for MOB6, and then the credit overdue probability of the target object is calculated based on the first probability weight, the first overdue probability, the second probability weight and the second overdue probability, which represents the credit overdue risk of the target object. For example, the first probability weight can be 0.3 and the second probability weight can be 0.7.

[0017] In order to make the joint risk prediction model have multi-task prediction ability, it needs to be trained in advance, Figure 2 is a flowchart of a model training method provided by an example embodiment of the present application, as shown in the figure, the method comprises: Figure 2 Step 201, obtaining sample credit data of a plurality of sample objects; Step 202, inputting the sample credit data into the long-short term shared layer for feature extraction to obtain sample long-term features and sample short-term features; Step 203, inputting the sample short-term features into the short-term prediction network to obtain the first sample probability output by the short-term prediction network; Step 204, inputting the sample long-term features into the long-term prediction network to obtain the second sample probability output by the long-term prediction network; Step 205, training the long-short term shared layer, the short-term prediction network and the long-term prediction network based on the first sample probability, the short-term overdue label, the second sample probability and the long-term overdue label.

[0018] Among them, the training sample of the model is the sample credit data and the overdue label of the plurality of sample objects, the sample credit data includes the education information, the income information, the historical transaction information, the historical credit information, the historical debt information and the age information of the sample object, and the overdue label includes the overdue label of MOB3 and the overdue label of MOB6.

[0019] Before the model training, in order to distinguish the influence of the sample features in the sample credit data on the long-short term overdue risk, through the causal discovery algorithm, based on the sample credit data and the overdue label, the differential driving features of MOB3 and MOB6+ are identified (such as MOB3 is affected by income fluctuation, MOB6+ is affected by debt structure), so as to determine the sample causal relationship of the long-short term overdue risk, and the sample causal relationship can be the influence weight of different sample features on the long-short term risk. ​In each round of training process, the sample credit data of each sample object is input into the long short-term shared layer for feature extraction to obtain sample long-term features and sample short-term features. Then, the sample short-term features are input into the short-term prediction network to obtain a first sample probability output by the short-term prediction network, the first sample probability representing a short-term delinquency risk of the sample object. The sample long-term features are input into the long-term prediction network to obtain a second sample probability output by the long-term prediction network, the second sample probability representing a long-term delinquency risk of the sample object. In order to enable the model to learn correct long short-term delinquency targets, based on the loss between the first sample probability and the short-term delinquency label and the loss between the second sample probability and the long-term prediction label, the long short-term shared layer, the short-term prediction network and the long-term prediction network are trained.

[0020] In order to avoid task conflicts, causal constraints are also introduced in the loss calculation of the model. Corresponding to an exemplary example, step 205 can include steps 205A to 205E.

[0021] Step 205A, determining a short-term prediction loss based on the first sample probability and the short-term delinquency label; Step 205B, determining a long-term prediction loss based on the second sample probability and the long-term delinquency label; Step 205C, determining a causal penalty term based on the sample short-term features, the sample long-term features and the sample causal relationship; Step 205D, determining a target total loss based on the short-term prediction loss, the first loss weight, the long-term prediction loss, the second loss weight, the causal penalty term and the third loss weight; Step 205E, training the long short-term shared layer, the short-term prediction network and the long-term prediction network based on the target total loss.

[0022] Exemplarily, the target total loss of the joint risk prediction model can be: Loss = a*Loss_MOB3 + β*Loss_MOB6+ + γ*Causal_Constraint; wherein, Loss_MOB3 is the short-term prediction loss, Loss_MOB6 is the long-term prediction loss, Causal_Constraint is the causal penalty term, a is the first loss weight of the short-term prediction loss, β is the second loss weight of the long-term prediction loss, and γ is the third loss weight of the causal penalty term. a can be 0.4, and β can be 0.6.

[0023] In a possible implementation, in each round of the training process, a short-term prediction loss is determined based on a difference between a first sample probability output by the short-term prediction network and a short-term overdue label, a long-term prediction loss is determined based on a difference between a second sample probability output by the long-term prediction network and a long-term overdue label, and a causal penalty term is determined based on a sample short-term feature, a sample long-term feature, and a sample causal relationship (an influence weight of different sample features on long-term and short-term risks inferred by a causal algorithm); and then a target total loss is determined based on the short-term prediction loss and a first loss weight, the long-term prediction loss and a second loss weight, and the causal penalty term and a third loss weight, so as to train the long-term and short-term shared layer, the short-term prediction network, and the long-term prediction network based on the target total loss.

[0024] Through multiple rounds of iterative training, until the target total loss is less than a preset value, or the number of iterations reaches a preset number, the model training is determined to be completed, and the trained long-term and short-term shared layer, the trained short-term prediction network, and the trained long-term prediction network are obtained.

[0025] In view of the difference in feature distribution between online (source domain) and offline (target domain), in order to improve the generalization of the model, the online model is migrated to the offline through transfer learning. In a possible implementation, the general feature representation is first pre-trained using online data to encode the sample credit data of each sample object; and then the model is fine-tuned using offline data.

[0026] In view of the fact that the sample credit data includes multiple sample features, and different sample features have different influences on long-term and short-term risk prediction, in order to make the model more easily learn the difference, the sample weight of the sample credit data is also dynamically adjusted. In an exemplary example, step 202 can include steps 202A to 202C.

[0027] Step 202A, for each sample object, determining a feature weight of each sample credit feature in the sample credit data; In the sample credit data, multiple sample credit features are included, the feature weight of each sample credit feature is set according to a preset rule, and the model is trained using the weighted sample credit data.

[0028] Specifically, determining the feature weight of each sample credit feature in the sample credit data can include: obtaining time information of each sample credit feature; and determining the feature weight of each sample credit feature in the sample credit data based on the time information, the feature weight and the time information being in a negative correlation relationship.

[0029] It is considered that the sample credit feature generated closer to the current time is more likely to affect the risk prediction result. Correspondingly, the time information of each sample credit feature is obtained, and the feature weight of each sample credit feature in the sample credit data is determined based on the difference between the time information and the current time. The feature weight and the proximity of the time information are negatively correlated. That is, the closer the time information indicates to the current time, the greater the feature weight, and vice versa. The farther the time information indicates from the current time, the relatively smaller the feature weight.

[0030] Optionally, determining the feature weight of each sample credit feature in the sample credit data can further include: determining the recovery rate of the sample object based on the sample credit data; and determining the feature weight of each sample credit feature in the sample credit data based on the recovery rate.

[0031] It is considered that there are sample objects with high recovery rate after MOB, and the short-term overdue of these sample objects can be repaired, so it is necessary to adjust the feature weight of the sample credit data corresponding to these sample objects. In one possible implementation, the recovery rate of the sample object is determined based on the sample credit data; if the recovery rate indicates that the sample object can be repaid through collection, the weight of the sample credit feature related to MOB3 in the sample credit data is reduced.

[0032] Step 20B, the sample credit data is weighted based on the feature weight to obtain weighted sample data; Step 202C, the weighted sample data is input into the long-short term shared layer for feature extraction to obtain sample long-term features and sample short-term features.

[0033] After obtaining the feature weight of each sample credit feature, the feature weight can be used to weight each sample credit feature to obtain weighted sample data after weighting. Then the weighted sample data is input into the long-short term shared layer for feature extraction to learn the influence degree of different sample credit features on long-short term risk prediction, and sample long-term features and sample short-term features are obtained.

[0034] In summary, the embodiment of the present application provides a credit data processing method: when it is necessary to predict the overdue risk of a target object, object credit data of the target object can be acquired, and the object credit data is input into a trained long-short-term shared layer for feature extraction to obtain target long-term features and target short-term features, then the target short-term features are input into a trained short-term prediction network to obtain a first overdue probability output by the short-term prediction network; and the target long-term features are input into a trained long-term prediction network to obtain a second overdue probability output by the long-term prediction network, and then the credit overdue risk of the target object is determined according to the first overdue probability and / or the second overdue probability; through multi-task joint prediction, the long-short-term overdue risk prediction of the target object can be realized, and the overdue prediction demand of different risk observation periods on the strategy side can be met while meeting the model prediction stability demand.

[0035] Please refer to Figure 3 , which is a structural schematic diagram of a credit data processing device provided by the embodiment of the present application. As shown in the example of Figure 3 , the device 300 comprises: A first acquisition module 301 is configured to acquire object credit data of a target object. A first feature extraction module 302 is configured to input the object credit data into a trained long-short-term shared layer for feature extraction to obtain target long-term features and target short-term features. A first prediction module 303 is configured to input the target short-term features into a trained short-term prediction network to obtain a first overdue probability output by the short-term prediction network. A second prediction module 304 is configured to input the target long-term features into a trained long-term prediction network to obtain a second overdue probability output by the long-term prediction network. A determination module 305 is configured to determine the credit overdue risk of the target object based on the first overdue probability and / or the second overdue probability.

[0036] Optionally, the device further comprises: A second acquisition module is configured to acquire sample credit data of a plurality of sample objects. A second feature extraction module is configured to input the sample credit data into a long-short-term shared layer for feature extraction to obtain sample long-term features and sample short-term features. A third prediction module is configured to input the sample short-term features into a short-term prediction network to obtain a first sample probability output by the short-term prediction network. A fourth prediction module is configured to input the sample long-term features into a long-term prediction network to obtain a second sample probability output by the long-term prediction network. a training module configured to train the long-short term shared layer, the short-term prediction network and the long-term prediction network based on the first sample probability, the short-term overdue label, the second sample probability and the long-term overdue label.

[0037] Optionally, the training module is further configured to: determine a short-term prediction loss based on the first sample probability and the short-term overdue label; determine a long-term prediction loss based on the second sample probability and the long-term overdue label; determine a causal penalty term based on the sample short-term feature, the sample long-term feature and a sample causal relationship; determine a target total loss based on the short-term prediction loss, a first loss weight, the long-term prediction loss, a second loss weight, the causal penalty term and a third loss weight; train the long-short term shared layer, the short-term prediction network and the long-term prediction network based on the target total loss.

[0038] Optionally, the sample credit data comprises a plurality of sample credit features. The second feature extraction module is further configured to: determine a feature weight of each sample credit feature in the sample credit data for each sample object; perform weighted processing on the sample credit data based on the feature weight to obtain weighted sample data; input the weighted sample data into the long-short term shared layer for feature extraction to obtain the sample long-term feature and the sample short-term feature.

[0039] Optionally, the second feature extraction module is further configured to: obtain time information of each sample credit feature; determine the feature weight of each sample credit feature in the sample credit data based on the time information, the feature weight being in a negative correlation with the time information.

[0040] Optionally, the second feature extraction module is further configured to: determine a recovery rate of the sample object based on the sample credit data; determine the feature weight of each sample credit feature in the sample credit data based on the recovery rate.

[0041] The exemplary embodiments of the present application further provide an electronic device, comprising: at least one processor; and a memory connected with the at least one processor in communication. The memory stores a computer program capable of being executed by the at least one processor, and the computer program, when executed by the at least one processor, is configured to cause the electronic device to perform the method according to the embodiments of the present application.

[0042] The exemplary embodiments of the present application further provide a non-transitory computer readable storage medium storing a computer program, wherein the computer program, when executed by a processor of a computer, is configured to cause the computer to perform the method according to the embodiments of the present application.

[0043] The exemplary embodiments of the present application further provide a computer program product comprising a computer program, wherein the computer program, when executed by a processor of a computer, is configured to cause the computer to perform the method according to the embodiments of the present application.

[0044] Reference Figure 4 A block diagram of the structure of an electronic device 400 that can be a server or a client of the present application, which is an example of a hardware device that can be applied to various aspects of the present application, will now be described. The electronic device is intended to represent a wide variety of digital electronic computer devices, such as laptops, desktops, workstations, personal digital assistants, servers, blade servers, mainframes, and other appropriate computers. The electronic device can also represent a wide variety of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions, are meant to be examples only, and are not meant to limit implementations of the present application described and / or claimed in this document.

[0045] As shown in Figure 4 The electronic device 400 includes a computing unit 401 that can perform various appropriate actions and processes in accordance with a computer program stored in a read-only memory (ROM) 402 or a computer program loaded from a storage unit 408 into a random access memory (RAM) 403. Various programs and data required for the operation of the electronic device 400 can also be stored in the RAM 403. The computing unit 401, the ROM 402, and the RAM 403 are connected to each other through a bus 404. An input / output (I / O) interface 405 is also connected to the bus 404.

[0046] Multiple components in electronic device 400 are connected to I / O interface 405, including: input unit 406, output unit 407, storage unit 408, and communication unit 409. Input unit 406 can be any type of device capable of inputting information to electronic device 400. Input unit 406 can receive input digital or character information and generate key signal inputs related to user settings and / or function control of electronic device. Output unit 407 can be any type of device capable of presenting information and may include, but is not limited to, a display, speaker, video / audio output terminal, vibrator, and / or printer. Storage unit 408 may include, but is not limited to, disks and optical discs. Communication unit 409 allows electronic device 400 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks, and may include, but is not limited to, modems, network cards, infrared communication devices, wireless communication transceivers, and / or chipsets, such as Bluetooth devices, WiFi devices, WiMax devices, cellular communication devices, and / or the like.

[0047] The computing unit 401 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 401 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 401 performs the various methods and processes described above. For example, in some embodiments, Figure 1 , Figure 2 The method shown can be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as storage unit 408. In some embodiments, part or all of the computer program can be loaded and / or installed on electronic device 400 via ROM 402 and / or communication unit 409. In some embodiments, computing unit 401 can be configured to execute by any other suitable means (e.g., by means of firmware). Figure 1 , Figure 2 The method shown.

[0048] The program code used to implement the methods of this application may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing device, such that when executed by the processor or controller, the functions / operations specified in the flowcharts and / or block diagrams are implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0049] In the context of this application, a machine-readable medium can be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine- readable medium can include but is not limited to an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of the machine-readable storage medium will include one or more lines of electrical connections, portable computer disks, hard disk drives, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or Flash memory), optical fibers, portable compact disc read-only memories (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0050] As used in this application, the terms "machine-readable medium" and "computer- readable medium" refer to any computer program product, apparatus and / or device (e.g., magnetic discs, optical disks, memory, Programmable Logic Devices (PLDs)) used to provide machine instructions and / or data to a programmable processor, including a machine-readable medium that receives machine instructions as a machine-readable signal. The term "machine-readable signal" refers to any signal that can be used to provide machine instructions and / or data to a programmable processor.

[0051] To provide for interaction with a user, the systems and techniques described here can be implemented on a computer having a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user and a keyboard and a pointing device (e.g., a mouse or a trackball) by which the user can provide input to the computer. Other kinds of devices can be used to provide for interaction with a user as well; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form, including acoustic, speech, or tactile input.

[0052] The systems and techniques described here can be implemented in a computing system that includes a back end component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front end component (e.g., a user computer having a graphical user interface or a Web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such back end, middleware, or front end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network (LAN), a wide area network (WAN), and the Internet.

[0053] The computing system can include clients and servers. A client and server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other.

Claims

1. A credit data processing method characterized by comprising: The method comprises: acquiring object credit data of a target object; inputting the object credit data into a trained long-short-term shared layer to perform feature extraction, to obtain target long-term features and target short-term features; inputting the target short-term features into a trained short-term prediction network to obtain a first overdue probability output by the short-term prediction network; inputting the target long-term features into a trained long-term prediction network to obtain a second overdue probability output by the long-term prediction network; based on the first overdue probability and / or the second overdue probability, determining a credit overdue risk of the target object.

2. The method of claim 1, wherein, The method further comprises: acquiring sample credit data of a plurality of sample objects; inputting the sample credit data into a long-short-term shared layer to perform feature extraction, to obtain sample long-term features and sample short-term features; inputting the sample short-term features into a short-term prediction network to obtain a first sample probability output by the short-term prediction network; inputting the sample long-term features into a long-term prediction network to obtain a second sample probability output by the long-term prediction network; based on the first sample probability, a short-term overdue label, the second sample probability, and a long-term overdue label, training the long-short-term shared layer, the short-term prediction network, and the long-term prediction network.

3. The method of claim 2, wherein, The training of the long-short-term shared layer, the short-term prediction network, and the long-term prediction network based on the first sample probability, the short-term overdue label, the second sample probability, and the long-term overdue label comprises: based on the first sample probability and the short-term overdue label, determining a short-term prediction loss; based on the second sample probability and the long-term overdue label, determining a long-term prediction loss; based on the sample short-term features, the sample long-term features, and a sample causal relationship, determining a causal penalty term; based on the short-term prediction loss, a first loss weight, the long-term prediction loss, a second loss weight, the causal penalty term, and a third loss weight, determining a target total loss; based on the target total loss, training the long-short-term shared layer, the short-term prediction network, and the long-term prediction network.

4. The method of claim 2, wherein, The sample credit data comprises a plurality of sample credit features; The inputting of the sample credit data into a long-short-term shared layer to perform feature extraction, to obtain sample long-term features and sample short-term features, comprises: for each of the sample objects, determining feature weights of respective sample credit features in the sample credit data; based on the feature weights, performing weighted processing on the sample credit data to obtain weighted sample data; inputting the weighted sample data into the long-short-term shared layer to perform feature extraction, to obtain the sample long-term features and the sample short-term features.

5. The method of claim 4, wherein, The determination of the feature weights of respective sample credit features in the sample credit data comprises: acquiring time information of respective sample credit features; based on the time information, determining the feature weights of respective sample credit features in the sample credit data, the feature weights being in a negative correlation with the time information.

6. The method of claim 4, wherein, The determination of the feature weights of respective sample credit features in the sample credit data comprises: based on the sample credit data, determining a recovery rate of the sample object; Determine the feature weight of each sample credit feature in the sample credit data based on the recovery rate.

7. A credit data processing apparatus characterized by comprising: The device comprises: A first acquisition module configured to acquire object credit data of a target object; A first feature extraction module configured to input the object credit data into a trained long-short term shared layer to perform feature extraction and obtain target long-term features and target short-term features; A first prediction module configured to input the target short-term features into a trained short-term prediction network to obtain a first delinquency probability output by the short-term prediction network; A second prediction module configured to input the target long-term features into a trained long-term prediction network to obtain a second delinquency probability output by the long-term prediction network; A determination module configured to determine a credit delinquency risk of the target object based on the first delinquency probability and / or the second delinquency probability.

8. The apparatus of claim 7, wherein, The device further comprises: A second acquisition module configured to acquire sample credit data of a plurality of sample objects; A second feature extraction module configured to input the sample credit data into a long-short term shared layer to perform feature extraction and obtain sample long-term features and sample short-term features; A third prediction module configured to input the sample short-term features into a short-term prediction network to obtain a first sample probability output by the short-term prediction network; A fourth prediction module configured to input the sample long-term features into a long-term prediction network to obtain a second sample probability output by the long-term prediction network; A training module configured to train the long-short term shared layer, the short-term prediction network and the long-term prediction network based on the first sample probability, a short-term delinquency label, the second sample probability and a long-term delinquency label.

9. An electronic device comprising: a processor; and a memory storing a program, wherein the program includes instructions that, when executed by the processor, cause the processor to perform the credit data processing method according to any one of claims 1-6.

10. A non-transitory computer readable storage medium having stored thereon computer instructions, wherein, The computer instructions are used to cause the computer to perform the credit data processing method according to any one of claims 1-6.

Images