User behavior prediction method and device, computer equipment and storage medium
By introducing two prediction models of early execution and overdue execution into the user behavior prediction method, the user's early execution and overdue execution behaviors are predicted, which solves the problem of incomplete prediction in the existing technology and improves the stability of financial resource leasing. .
Patent Information
- Application Number
- CN202411503689.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-25
- Publication Date
- 2025-01-24
AI Technical Summary
The existing user behavior prediction methods only predict whether the user returns resources overdue, reducing the comprehensiveness of user behavior prediction.
By obtaining the target transaction data corresponding to the source transaction of the target user, an early execution prediction model is used to determine whether the target user has early execution behavior of executing the source transaction corresponding to the target transaction in advance, and if the target user does not have early return behavior, use overdue The execution prediction model determines whether the target user has overdue execution behavior of overdue execution of the target transaction.
It improves the comprehensiveness of user behavior prediction and ensures the stability after lease of financial resources.
Smart Images
Figure CN119359422A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of artificial intelligence technology, and in particular to a method, apparatus, computer device and storage medium for predicting user behavior, which can be applied to the financial field and other technical fields. Background Art
[0002] With the continuous development of the financial field, the transaction volume of financial resource leasing under various institutions has also increased. In order to ensure the stability of financial resource leasing, user behavior prediction methods have emerged. Through the transaction processing data of users in the historical period, it is predicted whether the user can comply with the resource leasing regulations.
[0003] However, the existing user behavior prediction methods only predict whether the user will return the resources overdue, which reduces the comprehensiveness of user behavior prediction. Summary of the invention
[0004] Based on this, it is necessary to provide a user behavior prediction method, device, computer equipment and storage medium that can improve the comprehensiveness of user behavior prediction in response to the above technical problems.
[0005] In a first aspect, the present application provides a method for predicting user behavior. The method comprises:
[0006] Obtain the target transaction data corresponding to the source transaction of the target user;
[0007] Adopting the early execution prediction model, based on the target transaction data, it is determined whether the target user has the early execution behavior of executing the source transaction corresponding to the target transaction in advance; wherein the source transaction and the target transaction have a causal relationship;
[0008] In the case that the target user does not have any early return behavior, the overdue execution prediction model is used to determine whether the target user has overdue execution behavior of overdue execution of the target transaction based on the target transaction data.
[0009] In one embodiment, the early execution prediction model includes at least two cascaded feature extraction networks and a feature fusion network; different feature extraction networks have different feature extraction scales; accordingly, the early execution prediction model is used to determine whether the target user has an early execution behavior of executing the source transaction corresponding to the target transaction in advance according to the target transaction data, including:
[0010] The input data to be processed is subjected to feature extraction through different feature extraction networks to obtain corresponding target behavior features; wherein, the data to be processed of the first feature extraction network is the target transaction data; the data to be processed of the non-first feature extraction network is the output data of the previous feature extraction network; the target behavior features are subjected to feature fusion through the feature fusion network to obtain fused behavior features; according to the fused behavior features, it is determined whether the target user has the early execution behavior of executing the source transaction in advance corresponding to the target transaction.
[0011] In one embodiment, the feature extraction network includes a basic feature extraction network; accordingly, the input data to be processed is subjected to feature extraction through different feature extraction networks to obtain corresponding target behavior features, including:
[0012] The input data to be processed is subjected to feature extraction through different basic feature extraction networks to obtain the corresponding target behavior features.
[0013] In one embodiment, the feature extraction network includes a basic feature extraction network and a local feature extraction network; accordingly, the input data to be processed is subjected to feature extraction through different feature extraction networks to obtain corresponding target behavior features, including:
[0014] Through the basic feature extraction network, feature extraction is performed on the input data to be processed to obtain the corresponding initial behavior features; through the local feature extraction network, local feature extraction is performed on the initial behavior features to obtain the corresponding target behavior features.
[0015] In one embodiment, the early execution prediction model further includes a channel weight determination network; accordingly, the feature fusion network is used to fuse the target behavior features to obtain fused behavior features, including:
[0016] Through the feature fusion network, the corresponding target behavior features are weighted and fused according to the weights of each feature to obtain the fused behavior features corresponding to the target user.
[0017] In one embodiment, the early execution prediction model is trained in the following manner:
[0018] Obtain sample transaction data corresponding to the source transaction of the sample user; input the sample transaction data into the initial prediction model to obtain the prediction result of the sample user; determine the loss value according to the actual result and the prediction result of the sample user; use the loss value to adjust the network parameters in the initial prediction model to obtain the early execution prediction model.
[0019] In a second aspect, the present application also provides a user behavior prediction device. The device comprises:
[0020] A data acquisition module, used to acquire target transaction data corresponding to the source transaction of the target user;
[0021] The first prediction module is used to use an early execution prediction model to determine whether the target user has an early execution behavior of executing a source transaction corresponding to a target transaction in advance according to the target transaction data; wherein the source transaction and the target transaction have a causal relationship;
[0022] The second prediction module is used to use the overdue execution prediction model to determine whether the target user has overdue execution of the target transaction based on the target transaction data when the target user has no early return behavior.
[0023] 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:
[0024] Obtain the target transaction data corresponding to the source transaction of the target user;
[0025] Adopting the early execution prediction model, based on the target transaction data, it is determined whether the target user has the early execution behavior of executing the source transaction corresponding to the target transaction in advance; wherein the source transaction and the target transaction have a causal relationship;
[0026] In the case that the target user does not have any early return behavior, the overdue execution prediction model is used to determine whether the target user has overdue execution behavior of overdue execution of the target transaction based on the target transaction data.
[0027] 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:
[0028] Obtain the target transaction data corresponding to the source transaction of the target user;
[0029] Adopting the early execution prediction model, based on the target transaction data, it is determined whether the target user has the early execution behavior of executing the source transaction corresponding to the target transaction in advance; wherein the source transaction and the target transaction have a causal relationship;
[0030] In the case that the target user does not have any early return behavior, the overdue execution prediction model is used to determine whether the target user has overdue execution behavior of overdue execution of the target transaction based on the target transaction data.
[0031] 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:
[0032] Obtain the target transaction data corresponding to the source transaction of the target user;
[0033] Adopting the early execution prediction model, based on the target transaction data, it is determined whether the target user has the early execution behavior of executing the source transaction corresponding to the target transaction in advance; wherein the source transaction and the target transaction have a causal relationship;
[0034] In the case that the target user does not have any early return behavior, the overdue execution prediction model is used to determine whether the target user has overdue execution behavior of overdue execution of the target transaction based on the target transaction data.
[0035] The above-mentioned user behavior prediction method, device, computer equipment and storage medium, by adopting an early execution prediction model, determine whether the target user has an early execution behavior of executing the source transaction corresponding to the target transaction in advance according to the target transaction data; then, in the case that the target user does not have an early return behavior, the overdue execution prediction model is adopted to determine whether the target user has an overdue execution behavior of overdue execution of the target transaction according to the target transaction data. Compared with the related art, which only predicts whether the user has an overdue execution behavior, the above-mentioned method introduces the prediction of the early return behavior of the target user, which can ensure the comprehensiveness of the user prediction and thus ensure the stability of financial resources. BRIEF DESCRIPTION OF THE DRAWINGS
[0036] Figure 1 A diagram showing an application environment of a user behavior prediction method in one embodiment;
[0037] Figure 2 A schematic diagram of a flow chart of a user behavior prediction method in one embodiment;
[0038] Figure 3 A schematic diagram of a process for predicting early execution behavior in one embodiment;
[0039] Figure 4 is a flow chart of a method for predicting user behavior in another embodiment;
[0040] Figure 5 is a structural block diagram of a user behavior prediction device in one embodiment;
[0041] Figure 6 is a structural block diagram of a user behavior prediction device in another embodiment;
[0042] Figure 7 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION
[0043] 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.
[0044] The user behavior prediction method provided in the embodiment of the present application can be applied to Figure 1 In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. For example, the transaction data of the user. The data storage system can be integrated on the server 104, or it can be placed on the cloud or other network servers. For example, after the server 104 obtains the target transaction data corresponding to the source transaction of the target user, it adopts the early execution prediction model to determine whether the target user has an early execution behavior of executing the source transaction corresponding to the target transaction in advance according to the target transaction data. Subsequently, if the target user does not have an early return behavior, the overdue execution prediction model is adopted to determine whether the target user has an overdue execution behavior of overdue execution of the target transaction according to the target transaction data; further, the server 104 can send the predicted user behavior to the terminal 102 for display. Among them, the terminal 102 can be but not limited to various personal computers, laptops, smart phones, tablet computers, Internet of Things devices, etc. The server 104 can be implemented with an independent server or a server cluster consisting of multiple servers.
[0045] With the continuous development of the financial field, the transaction volume of financial resource leasing under various institutions has also increased. In order to ensure the stability of financial resource leasing, user behavior prediction methods have emerged. Through the transaction processing data of users in the historical period, it is predicted whether the user can comply with the resource leasing regulations.
[0046] However, the existing user behavior prediction methods only predict whether the user will return the resources overdue, which reduces the comprehensiveness of user behavior prediction.
[0047] Based on this, in one embodiment, Figure 2 As shown, a user behavior prediction method is provided, which is applied to Figure 1 The server in the example is used as an example to illustrate the specific steps:
[0048] S201, obtaining target transaction data corresponding to a source transaction of a target user.
[0049] Among them, the target user refers to the user under the target organization; the source transaction refers to the transaction related to resource transfer behavior within the target organization, for example, the pre-resource leasing transaction; the target transaction data refers to the behavior data generated when the target user performs operations related to the source transaction.
[0050] Optionally, when predicting the behavior of the target user, the user identifier of the target user (for example, user account information) can be entered on the terminal and click Send to generate a behavior prediction request carrying the user identifier and send it to the server; further, after the server receives the behavior prediction request, it can obtain the target transaction data corresponding to the source transaction of the target user based on the user identifier in the behavior prediction request.
[0051] It is understandable that in order to ensure the accuracy and validity of the target transaction data, the unprocessed original transaction data corresponding to the source transaction of the target user may be obtained first; then, the original transaction data may be preprocessed to obtain the target transaction data.
[0052] Exemplarily, missing data and duplicate data in the original transaction data may be deleted first; then, the original transaction data remaining after the deletion process may be normalized to obtain the target transaction data.
[0053] S202, using an early execution prediction model, according to the target transaction data, determining whether the target user has an early execution behavior of executing the source transaction in advance corresponding to the target transaction.
[0054] Among them, the early execution prediction model is a trained neural network model, which can be used to predict whether the target user has early execution behavior; the source transaction and the target transaction have a causal relationship. For example, if the source transaction is a resource leasing transaction, the target transaction is a resource return transaction. At this time, the early execution behavior can be the early return of resources.
[0055] Optionally, the target transaction data may be input into an early execution prediction model, and the early execution prediction model determines whether the target user has an early execution behavior of executing the source transaction corresponding to the target transaction in advance based on the target transaction data and model parameters.
[0056] S203, when the target user does not have any early return behavior, an overdue execution prediction model is used to determine whether the target user has any overdue execution behavior of overdue execution of the target transaction according to the target transaction data.
[0057] Among them, the overdue execution prediction model is a trained neural network model, which can be used to predict whether the target user has overdue execution behavior. For example, when the target transaction is a resource return transaction, the overdue execution behavior may be overdue return of resources.
[0058] Optionally, if the early execution prediction model determines that the target user has early return behavior, the behavior prediction process of the target user is ended; if the early execution prediction model determines that the target user does not have early return behavior, the target transaction data is input into the overdue execution prediction model, and the overdue execution prediction model determines whether the target user has overdue execution behavior of the target transaction corresponding to the overdue execution source transaction based on the target transaction data and model parameters.
[0059] Furthermore, after the predicted behavior of the target user is determined, the prediction result can be sent to the operation and maintenance end, which performs subsequent processing based on the prediction result of the target user.
[0060] In the above user behavior prediction method, by adopting the early execution prediction model, according to the target transaction data, it is determined whether the target user has the early execution behavior of executing the source transaction corresponding to the target transaction in advance; then, in the case that the target user does not have the early return behavior, the overdue execution prediction model is adopted to determine whether the target user has the overdue execution behavior of overdue execution of the target transaction according to the target transaction data. Compared with the related art, which only predicts whether the user has the overdue execution behavior, the above method introduces the prediction of the early return behavior of the target user, which can ensure the comprehensiveness of the user prediction and thus ensure the stability of financial resources.
[0061] In order to ensure the accuracy of the prediction results, based on the above embodiment, in this embodiment, the early execution prediction model includes at least two cascaded feature extraction networks and a feature fusion network, and the feature extraction scales of different feature extraction networks are different; further, based on the above structure, an optional method for predicting early execution behavior is provided, such as Figure 3 As shown, the specific steps include:
[0062] S301, extracting features from the input data to be processed through different feature extraction networks to obtain corresponding target behavior features.
[0063] Among them, the feature extraction network refers to a neural network that can extract features from target transaction data; the target behavior feature refers to the behavior feature of the target user extracted from the target transaction data; further, the data to be processed by the first feature extraction network is the target transaction data, and the data to be processed by the non-first feature extraction network is the output data of the previous feature extraction network.
[0064] An optional method is that, when the feature extraction network only includes a basic feature extraction network, different basic feature extraction networks can be used to perform feature extraction on the input data to be processed to obtain corresponding target behavior features.
[0065] Optionally, for the first basic feature extraction network, the target transaction data can be used as the input of the feature extraction network, and after feature extraction is performed on the target transaction data, the target behavior features corresponding to the basic feature extraction network can be obtained; for any basic feature extraction network other than the first basic feature extraction network, the output of the previous basic feature extraction network cascaded with the basic feature extraction network can be used as the input of the basic feature extraction network, and then, after feature extraction is performed on the output of the previous basic feature extraction network, the target behavior features corresponding to the basic feature extraction network can be obtained.
[0066] For example, after the first basic feature extraction network A1 extracts features from the target transaction data, the target behavior feature A1' corresponding to the first basic feature extraction network A1 is obtained; further, the second basic feature extraction network A2 is used to extract features from the target behavior feature A1' to obtain the target behavior feature A2' corresponding to the second basic feature extraction network A2.
[0067] Another optional method is that, when the feature extraction network includes a basic feature extraction network and a local feature extraction network, the basic feature extraction network is used to extract features from the input data to be processed to obtain the corresponding initial behavior features, and the local feature extraction network is used to extract local features from the initial behavior features to obtain the corresponding target behavior features. The local feature extraction network refers to a network that extracts features from a part of the initial behavior features.
[0068] Optionally, for the first basic feature extraction network, the target transaction data can be used as the input of the feature extraction network. After feature extraction of the target transaction data, the obtained initial behavior features can be input into the local feature extraction network corresponding to the first basic feature extraction network to perform local feature extraction, thereby obtaining the target behavior features corresponding to the basic feature extraction network.
[0069] Furthermore, for any basic feature extraction network other than the first one, the output of the previous basic feature extraction network cascaded with the basic feature extraction network can be used as the input of the basic feature extraction network. Subsequently, after feature extraction is performed on the output of the previous basic feature extraction network, the obtained initial behavior features can be input into the local feature extraction network corresponding to the basic feature extraction network to perform local feature extraction, thereby obtaining the target behavior features corresponding to the basic feature extraction network.
[0070] S302, performing feature fusion on each target behavior feature through a feature fusion network to obtain a fused behavior feature.
[0071] Among them, the feature fusion network refers to a neural network that fuses the target behavior features output by each feature extraction network; the fused behavior feature refers to the behavior feature obtained by fusing each target behavior feature.
[0072] Optionally, the target behavior features output by each feature extraction network can be input into the feature fusion network for fusion processing to obtain the fused behavior features corresponding to the target user. Exemplarily, in order to ensure the accuracy of the fused behavior features, the feature fusion network can be used to perform weighted fusion on the corresponding target behavior features according to the weights of each feature to obtain the fused behavior features corresponding to the target user.
[0073] Optionally, the feature weights corresponding to each feature extraction network can be determined according to the importance of each feature extraction network; then, the feature fusion network can perform weighted fusion on the corresponding target behavior features according to the feature weights corresponding to each feature extraction network, so as to obtain the fused behavior features corresponding to the target user.
[0074] S303: Determine, based on the fusion behavior feature, whether the target user has an early execution behavior of executing the source transaction corresponding to the target transaction in advance.
[0075] Optionally, a classifier may be used to determine whether the target user has an early execution behavior of executing the source transaction corresponding to the target transaction in advance based on the fused behavior characteristics. For example, a random forest algorithm may be used to construct a corresponding decision tree based on the fused behavior characteristics, thereby determining whether the target user has an early execution behavior of executing the source transaction corresponding to the target transaction in advance.
[0076] It is understandable that when predicting the overdue execution behavior of the target user, the above steps S301-S303 may also be referred to for prediction.
[0077] In this embodiment, a feature extraction network and a feature fusion network are introduced. The target transaction data is processed by the feature extraction network and the feature fusion network to ensure the accuracy of the prediction results.
[0078] In order to ensure the accuracy of model training, based on the above embodiments, in this embodiment, an optional method of early execution prediction model training is provided. It can be understood that the training process of early execution prediction model and overdue execution prediction model is the same. Specifically, the sample transaction data corresponding to the source transaction of the sample user is obtained, and the sample transaction data is input into the initial prediction model to obtain the prediction result of the sample user; the loss value is determined according to the actual result and the prediction result of the sample user; then, the network parameters in the initial prediction model are adjusted using the loss value to obtain the early execution prediction model.
[0079] Optionally, the sample transaction data corresponding to the source transaction of the sample user can be input into the initial prediction model, and the initial prediction model determines the prediction result based on the sample transaction data and model parameters; then, a preset loss function can be used to determine the loss value based on the actual result and the predicted result of the sample user; finally, the network parameters in the initial prediction model can be adjusted according to the loss value, and when the accuracy of the initial prediction model reaches a preset threshold, an early execution prediction model can be obtained.
[0080] In this embodiment, an optional method of model training is provided, which can ensure the accuracy of model training.
[0081] Figure 4 FIG. 2 is a flow chart of a method for predicting user behavior in another embodiment. Based on the above embodiment, this embodiment provides an optional example of a method for predicting user behavior. Figure 4 The specific implementation process is as follows:
[0082] S401, obtaining target transaction data corresponding to a source transaction of a target user.
[0083] S402, using the advance execution prediction model, extracting features from the input data to be processed through different feature extraction networks to obtain corresponding target behavior features.
[0084] Among them, the data to be processed by the first feature extraction network is the target transaction data; the data to be processed by the non-first feature extraction network is the output data of the previous feature extraction network.
[0085] Optionally, when the feature extraction network includes a basic feature extraction network and a local feature extraction network, feature extraction is performed on the input data to be processed through the basic feature extraction network to obtain corresponding initial behavior features; and local feature extraction is performed on the initial behavior features through the local feature extraction network to obtain corresponding target behavior features.
[0086] S403, through the feature fusion network, weighted fusion is performed on the corresponding target behavior features according to the weights of each feature, so as to obtain the fused behavior features corresponding to the target user.
[0087] S404: Determine, based on the fusion behavior feature, whether the target user has an early execution behavior of executing the source transaction corresponding to the target transaction in advance.
[0088] Among them, the source transaction and the target transaction have a causal relationship;
[0089] S405, when the target user does not have any early return behavior, the overdue execution prediction model is used to extract features of the input data to be processed through different feature extraction networks to obtain corresponding target behavior features.
[0090] S406, performing feature fusion on each target behavior feature through a feature fusion network to obtain a fused behavior feature.
[0091] S407, determining whether the target user has an overdue execution behavior of the target transaction corresponding to the overdue execution source transaction according to the fusion behavior characteristics.
[0092] Optionally, sample transaction data corresponding to the source transaction of the sample user can be obtained; the sample transaction data is input into the initial prediction model to obtain the prediction result of the sample user; the loss value is determined according to the actual result and the prediction result of the sample user; the network parameters in the initial prediction model are adjusted using the loss value to obtain the early execution prediction model.
[0093] The specific process of the above S401-S407 can refer to the description of the above method embodiment, and its implementation principle and technical effect are similar, which will not be repeated here.
[0094] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned 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 does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can 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 to be 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.
[0095] Based on the same inventive concept, the embodiment of the present application also provides a user behavior prediction device for implementing the user behavior prediction method 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 one or more user behavior prediction device embodiments provided below can refer to the limitations of the user behavior prediction method above, and will not be repeated here.
[0096] In one embodiment, Figure 5 As shown, a user behavior prediction device 1 is provided, comprising: a data acquisition module 10, a first prediction module 20 and a second prediction module 30, wherein:
[0097] A data acquisition module 10 is used to acquire target transaction data corresponding to the source transaction of the target user;
[0098] The first prediction module 20 is used to use an early execution prediction model to determine whether the target user has an early execution behavior of executing a source transaction corresponding to a target transaction in advance according to the target transaction data; wherein the source transaction and the target transaction have a causal relationship;
[0099] The second prediction module 30 is used to use the overdue execution prediction model to determine whether the target user has overdue execution of the target transaction according to the target transaction data when the target user has no early return behavior.
[0100] In one embodiment, the advance execution prediction model includes at least two cascaded feature extraction networks and a feature fusion network; different feature extraction networks have different feature extraction scales; accordingly, Figure 6 As shown, the first prediction module 20 includes:
[0101] The feature extraction unit 21 is used to extract features from the input data to be processed through different feature extraction networks to obtain corresponding target behavior features; wherein the data to be processed by the first feature extraction network is the target transaction data; the data to be processed by the non-first feature extraction network is the output data of the previous feature extraction network;
[0102] A feature fusion unit 22 is used to fuse the target behavior features through a feature fusion network to obtain a fused behavior feature;
[0103] The prediction unit 23 is used to determine whether the target user has an early execution behavior of executing the target transaction corresponding to the source transaction in advance according to the fused behavior characteristics.
[0104] In one embodiment, the feature extraction network includes a basic feature extraction network; accordingly, the feature extraction unit 21 is specifically used for:
[0105] The input data to be processed is subjected to feature extraction through different basic feature extraction networks to obtain the corresponding target behavior features.
[0106] In one embodiment, the feature extraction network includes a basic feature extraction network and a local feature extraction network; accordingly, the feature extraction unit 21 is also used for:
[0107] Through the basic feature extraction network, the input data to be processed is subjected to feature extraction to obtain the corresponding initial behavior features; through the local feature extraction network, the initial behavior features are subjected to local feature extraction to obtain the corresponding target behavior features.
[0108] In one embodiment, the advance execution prediction model further includes a channel weight determination network; accordingly, the feature fusion unit 22 is specifically used for:
[0109] Through the feature fusion network, the corresponding target behavior features are weighted and fused according to the weights of each feature to obtain the fused behavior features corresponding to the target user.
[0110] In one embodiment, the early execution prediction model is trained in the following manner:
[0111] Obtain sample transaction data corresponding to the source transaction of the sample user; input the sample transaction data into the initial prediction model to obtain the prediction result of the sample user; determine the loss value according to the actual result and the prediction result of the sample user; use the loss value to adjust the network parameters in the initial prediction model to obtain the early execution prediction model.
[0112] Each module in the above user behavior prediction device can be implemented in whole or in part by software, hardware, or a combination thereof. Each module 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 module.
[0113] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Figure 7 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. 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, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store user transaction data. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a user behavior prediction method is implemented.
[0114] Those skilled in the art will understand that Figure 7 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.
[0115] 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:
[0116] Obtain the target transaction data corresponding to the source transaction of the target user;
[0117] Adopting the early execution prediction model, based on the target transaction data, it is determined whether the target user has the early execution behavior of executing the source transaction corresponding to the target transaction in advance; wherein the source transaction and the target transaction have a causal relationship;
[0118] In the case that the target user does not have any early return behavior, the overdue execution prediction model is used to determine whether the target user has overdue execution behavior of overdue execution of the target transaction based on the target transaction data.
[0119] In one embodiment, the early execution prediction model includes at least two cascaded feature extraction networks and a feature fusion network; different feature extraction networks have different feature extraction scales; when the processor executes the early execution prediction model in the computer program, and determines whether the target user has the logic of early execution behavior of early execution of the source transaction corresponding to the target transaction according to the target transaction data, the following steps are specifically implemented:
[0120] The input data to be processed is subjected to feature extraction through different feature extraction networks to obtain corresponding target behavior features; wherein, the data to be processed of the first feature extraction network is the target transaction data; the data to be processed of the non-first feature extraction network is the output data of the previous feature extraction network; the target behavior features are subjected to feature fusion through the feature fusion network to obtain fused behavior features; according to the fused behavior features, it is determined whether the target user has the early execution behavior of executing the source transaction in advance corresponding to the target transaction.
[0121] In one embodiment, the feature extraction network includes a basic feature extraction network; when the processor executes the logic of extracting features of the input data to be processed through different feature extraction networks in the computer program to obtain corresponding target behavior features, the following steps are specifically implemented:
[0122] The input data to be processed is subjected to feature extraction through different basic feature extraction networks to obtain the corresponding target behavior features.
[0123] In one embodiment, the feature extraction network includes a basic feature extraction network and a local feature extraction network; when the processor executes the logic of extracting features of the input data to be processed through different feature extraction networks in the computer program to obtain corresponding target behavior features, the following steps are specifically implemented:
[0124] Through the basic feature extraction network, feature extraction is performed on the input data to be processed to obtain the corresponding initial behavior features; through the local feature extraction network, local feature extraction is performed on the initial behavior features to obtain the corresponding target behavior features.
[0125] In one embodiment, the advance execution prediction model further includes a channel weight determination network; when the processor executes the logic of performing feature fusion on each target behavior feature through the feature fusion network in the computer program to obtain the fused behavior feature, the following steps are specifically implemented:
[0126] Through the feature fusion network, the corresponding target behavior features are weighted and fused according to the weights of each feature to obtain the fused behavior features corresponding to the target user.
[0127] In one embodiment, the early execution prediction model is trained in the following manner:
[0128] Obtain sample transaction data corresponding to the source transaction of the sample user; input the sample transaction data into the initial prediction model to obtain the prediction result of the sample user; determine the loss value according to the actual result and the prediction result of the sample user; use the loss value to adjust the network parameters in the initial prediction model to obtain the early execution prediction model.
[0129] 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:
[0130] Obtain the target transaction data corresponding to the source transaction of the target user;
[0131] Adopting the early execution prediction model, based on the target transaction data, it is determined whether the target user has the early execution behavior of executing the source transaction corresponding to the target transaction in advance; wherein the source transaction and the target transaction have a causal relationship;
[0132] In the case that the target user does not have any early return behavior, the overdue execution prediction model is used to determine whether the target user has overdue execution behavior of overdue execution of the target transaction based on the target transaction data.
[0133] In one embodiment, the early execution prediction model includes at least two cascaded feature extraction networks and a feature fusion network; different feature extraction networks have different feature extraction scales; when the code logic of using the early execution prediction model in a computer program to determine whether a target user has an early execution behavior of executing a source transaction corresponding to a target transaction in advance according to target transaction data is executed by a processor, the following steps are specifically implemented:
[0134] The input data to be processed is subjected to feature extraction through different feature extraction networks to obtain corresponding target behavior features; wherein, the data to be processed of the first feature extraction network is the target transaction data; the data to be processed of the non-first feature extraction network is the output data of the previous feature extraction network; the target behavior features are subjected to feature fusion through the feature fusion network to obtain fused behavior features; according to the fused behavior features, it is determined whether the target user has the early execution behavior of executing the source transaction in advance corresponding to the target transaction.
[0135] In one embodiment, the feature extraction network includes a basic feature extraction network; when the code logic of extracting features of the input data to be processed by different feature extraction networks to obtain corresponding target behavior features in the computer program is executed by the processor, the following steps are specifically implemented:
[0136] The input data to be processed is subjected to feature extraction through different basic feature extraction networks to obtain the corresponding target behavior features.
[0137] In one embodiment, the feature extraction network includes a basic feature extraction network and a local feature extraction network; when the code logic of extracting features of the input data to be processed by different feature extraction networks to obtain corresponding target behavior features in the computer program is executed by the processor, the following steps are specifically implemented:
[0138] Through the basic feature extraction network, the input data to be processed is subjected to feature extraction to obtain the corresponding initial behavior features; through the local feature extraction network, the initial behavior features are subjected to local feature extraction to obtain the corresponding target behavior features.
[0139] In one embodiment, the advance execution prediction model further includes a channel weight determination network; in the computer program, the feature fusion network is used to fuse the target behavior features, and when the code logic of obtaining the fused behavior features is executed by the processor, the following steps are specifically implemented:
[0140] Through the feature fusion network, the corresponding target behavior features are weighted and fused according to the weights of each feature to obtain the fused behavior features corresponding to the target user.
[0141] In one embodiment, the early execution prediction model is trained in the following manner:
[0142] Obtain sample transaction data corresponding to the source transaction of the sample user; input the sample transaction data into the initial prediction model to obtain the prediction result of the sample user; determine the loss value according to the actual result and the prediction result of the sample user; use the loss value to adjust the network parameters in the initial prediction model to obtain the early execution prediction model.
[0143] In one embodiment, a computer program product is provided, comprising a computer program, which, when executed by a processor, implements the following steps:
[0144] Obtain the target transaction data corresponding to the source transaction of the target user;
[0145] Adopting the early execution prediction model, based on the target transaction data, it is determined whether the target user has the early execution behavior of executing the source transaction corresponding to the target transaction in advance; wherein the source transaction and the target transaction have a causal relationship;
[0146] In the case that the target user does not have any early return behavior, the overdue execution prediction model is used to determine whether the target user has overdue execution behavior of overdue execution of the target transaction based on the target transaction data.
[0147] In one embodiment, the early execution prediction model includes at least two cascaded feature extraction networks and a feature fusion network; different feature extraction networks have different feature extraction scales; when the computer program is executed by the processor using the early execution prediction model, and determining whether the target user has an operation of early execution of the source transaction corresponding to the target transaction according to the target transaction data, the following steps are specifically implemented:
[0148] The input data to be processed is subjected to feature extraction through different feature extraction networks to obtain corresponding target behavior features; wherein, the data to be processed of the first feature extraction network is the target transaction data; the data to be processed of the non-first feature extraction network is the output data of the previous feature extraction network; the target behavior features are subjected to feature fusion through the feature fusion network to obtain fused behavior features; according to the fused behavior features, it is determined whether the target user has the early execution behavior of executing the source transaction in advance corresponding to the target transaction.
[0149] In one embodiment, the feature extraction network includes a basic feature extraction network; when the computer program is executed by the processor to extract features from the input data to be processed through different feature extraction networks to obtain the corresponding target behavior features, the following steps are specifically implemented:
[0150] The input data to be processed is subjected to feature extraction through different basic feature extraction networks to obtain the corresponding target behavior features.
[0151] In one embodiment, the feature extraction network includes a basic feature extraction network and a local feature extraction network; when the computer program is executed by the processor to extract features from the input data to be processed through different feature extraction networks to obtain the corresponding target behavior features, the following steps are specifically implemented:
[0152] Through the basic feature extraction network, the input data to be processed is subjected to feature extraction to obtain the corresponding initial behavior features; through the local feature extraction network, the initial behavior features are subjected to local feature extraction to obtain the corresponding target behavior features.
[0153] In one embodiment, the advance execution prediction model further includes a channel weight determination network; when the computer program is executed by the processor to fuse the target behavior features through the feature fusion network to obtain the fused behavior features, the following steps are specifically implemented:
[0154] Through the feature fusion network, the corresponding target behavior features are weighted and fused according to the weights of each feature to obtain the fused behavior features corresponding to the target user.
[0155] In one embodiment, the early execution prediction model is trained in the following manner:
[0156] Obtain sample transaction data corresponding to the source transaction of the sample user; input the sample transaction data into the initial prediction model to obtain the prediction result of the sample user; determine the loss value according to the actual result and the prediction result of the sample user; use the loss value to adjust the network parameters in the initial prediction model to obtain the early execution prediction model.
[0157] 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, and the collection, use and processing of relevant data must comply with relevant regulations.
[0158] 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 this 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.
[0159] 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.
[0160] 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 user behavior prediction method, characterized in that: The method comprises: Obtain the target transaction data corresponding to the source transaction of the target user; Adopting an early execution prediction model, according to the target transaction data, determining whether the target user has an early execution behavior of executing the target transaction corresponding to the source transaction in advance; wherein the source transaction and the target transaction have a causal relationship; In the case that the target user does not have any early return behavior, an overdue execution prediction model is used to determine whether the target user has an overdue execution behavior of overdue execution of the target transaction according to the target transaction data.
2. The method according to claim 1, characterized in that The advance execution prediction model includes at least two cascaded feature extraction networks and a feature fusion network; different feature extraction networks have different feature extraction scales; Accordingly, the adopting of the early execution prediction model to determine, according to the target transaction data, whether the target user has an early execution behavior of executing the target transaction corresponding to the source transaction in advance includes: The input data to be processed is subjected to feature extraction through different feature extraction networks to obtain corresponding target behavior features; wherein the data to be processed by the first feature extraction network is the target transaction data; the data to be processed by the non-first feature extraction network is the output data of the previous feature extraction network; Performing feature fusion on each of the target behavior features through a feature fusion network to obtain a fused behavior feature; According to the fusion behavior feature, it is determined whether the target user has an early execution behavior of executing the target transaction corresponding to the source transaction in advance.
3. The method according to claim 2, characterized in that The feature extraction network includes a basic feature extraction network; Accordingly, the input data to be processed is subjected to feature extraction through different feature extraction networks to obtain corresponding target behavior features, including: The input data to be processed is subjected to feature extraction through different basic feature extraction networks to obtain the corresponding target behavior features.
4. The method according to claim 2, characterized in that: The feature extraction network includes a basic feature extraction network and a local feature extraction network; Accordingly, the input data to be processed is subjected to feature extraction through different feature extraction networks to obtain corresponding target behavior features, including: Through the basic feature extraction network, feature extraction is performed on the input data to be processed to obtain corresponding initial behavior features; The local feature extraction network is used to extract local features from the initial behavior features to obtain corresponding target behavior features.
5. The method according to claim 4, characterized in that The advance execution prediction model also includes a channel weight determination network; Accordingly, the feature fusion network is used to fuse the target behavior features to obtain fused behavior features, including: Through the feature fusion network, the corresponding target behavior features are weighted and fused according to the weights of each feature to obtain the fused behavior features corresponding to the target user.
6. The method according to any one of claims 1 to 5, characterized in that: The advance execution prediction model is trained in the following way: Obtain sample transaction data corresponding to source transactions of sample users; Inputting the sample transaction data into the initial prediction model to obtain the prediction results of the sample users; Determine the loss value according to the actual results and the predicted results of the sample users; The loss value is used to adjust the network parameters in the initial prediction model to obtain an early execution prediction model.
7. A user behavior prediction device, characterized in that: The device comprises: A data acquisition module, used to acquire target transaction data corresponding to the source transaction of the target user; A first prediction module is used to use an early execution prediction model to determine whether the target user has an early execution behavior of executing the target transaction corresponding to the source transaction in advance according to the target transaction data; wherein the source transaction and the target transaction have a causal relationship; The second prediction module is used to use an overdue execution prediction model to determine whether the target user has overdue execution behavior of overdue execution of the target transaction according to the target transaction data when the target user has no early return behavior.
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 6 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 6 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 6 are implemented.