A method, device and apparatus for determining user credit

Abstract

The embodiments of this specification provide a method, device and equipment for determining user credit, which relates to the field of artificial intelligence technology, wherein the method includes: obtaining the behavior information of the target user from multiple data sources; merging the behavior information of the target user in multiple data sources according to key elements to obtain the behavior feature information set of the target user; based on the behavior feature information set of the target user, using the target prediction model to determine the characterization value of the target user; the target prediction model is a model for determining the user's credit according to the user's behavior characteristics obtained by combining the particle swarm algorithm, the feedforward neural network and the Adam optimization algorithm, and the characterization value is used to represent the target user's credit. In the embodiments of this specification, due to the use of the target prediction model obtained by combining the particle swarm algorithm, the feedforward neural network and the Adam optimization algorithm, the user's credit can be efficiently and accurately determined based on the user's behavior characteristics.

Description

Technical Field

[0001] The embodiments of this specification relate to the field of artificial intelligence technology, and in particular to a method, device and apparatus for determining user credit. Background Art

[0002] Since the outbreak of the subprime mortgage crisis, credit risk management has become the primary focus of financial institutions. In the prior art, banks and financial institutions usually rely on relatively fixed decision-making rules and human judgment to assess the credit risk of customers. This decision-making by expert system is somewhat subjective and is not enough to comprehensively and objectively assess the potential credit risk of customers. It can be seen that the technical solutions in the prior art cannot accurately determine the creditworthiness of users.

[0003] To address the above problems, no effective solution has been proposed yet. Summary of the invention

[0004] The embodiments of this specification provide a method, device and apparatus for determining user credit, so as to solve the problem that the user credit cannot be accurately determined in the prior art.

[0005] The embodiments of the present specification provide a method for determining user credit, comprising: obtaining behavior information of a target user from multiple data sources; merging the behavior information of the target user in the multiple data sources according to key elements to obtain a behavior feature information set of the target user; based on the behavior feature information set of the target user, determining a characterization value of the target user using a target prediction model; wherein the target prediction model is a model obtained by combining a particle swarm algorithm, a feedforward neural network and an Adam optimization algorithm for training to determine the user's credit according to the user's behavior features, and the characterization value is used to represent the target user's credit.

[0006] The embodiments of the present specification also provide a device for determining user credit, including: an acquisition module, used to obtain behavior information of a target user from multiple data sources; a merging module, used to merge the behavior information of the target user in the multiple data sources according to key elements to obtain a behavior feature information set of the target user; a determination module, used to determine a characterization value of the target user based on the behavior feature information set of the target user using a target prediction model; wherein the target prediction model is a model obtained by training in combination with a particle swarm algorithm, a feedforward neural network and an Adam optimization algorithm for determining a user's credit according to the user's behavior characteristics, and the characterization value is used to represent the credit of the target user.

[0007] The embodiment of the present specification also provides a device for determining user credit, including a processor and a memory for storing processor executable instructions, and the processor implements the steps of the method for determining user credit described in the embodiment of the present specification when executing the instructions.

[0008] The embodiments of the present specification also provide a computer-readable storage medium on which computer instructions are stored. When the instructions are executed, the steps of the method for determining the user's credit rating described in the embodiments of the present specification are implemented.

[0009] The embodiments of this specification provide a method for determining the user's credit, which can obtain the behavior information of the target user from multiple data sources, and merge the behavior information of the target user in the multiple data sources according to key factors to obtain the behavior feature information set of the target user. Furthermore, based on the behavior feature information set of the target user, a characterization value of the target user can be determined using a target prediction model trained by combining a particle swarm algorithm, a feedforward neural network and an Adam optimization algorithm, wherein the above-mentioned characterization value is used to represent the credit of the target user. Since a target prediction model trained by combining a particle swarm algorithm, a feedforward neural network and an Adam optimization algorithm is adopted, the user's credit can be determined efficiently and accurately based on the user's behavior characteristics. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] The drawings described herein are used to provide a further understanding of the embodiments of this specification, constitute a part of the embodiments of this specification, and do not constitute a limitation of the embodiments of this specification. In the drawings:

[0011] Figure 1 It is a schematic diagram of the steps of a method for determining user credit according to an embodiment of this specification;

[0012] Figure 2 is a schematic diagram of a process of performing equalization processing according to an embodiment of this specification;

[0013] Figure 3 It is a structural schematic diagram of a device for determining user credit according to an embodiment of this specification;

[0014] Figure 4 It is a structural schematic diagram of a device for determining user credit provided in an embodiment of this specification. DETAILED DESCRIPTION

[0015] The principles and spirit of the embodiments of this specification will be described below with reference to several exemplary embodiments. It should be understood that these embodiments are provided only to enable those skilled in the art to better understand and implement the embodiments of this specification, and are not intended to limit the scope of the embodiments of this specification in any way. On the contrary, these embodiments are provided to make the disclosure of the embodiments of this specification more thorough and complete, and to fully convey the scope of the disclosure to those skilled in the art.

[0016] Those skilled in the art know that the implementation of the embodiments of this specification can be implemented as a system, device, method or computer program product. Therefore, the embodiments disclosed in this specification can be specifically implemented in the following forms, namely: complete hardware, complete software (including firmware, resident software, microcode, etc.), or a combination of hardware and software.

[0017] Although the processes described below include multiple operations that occur in a particular order, it should be clear that these processes may include more or fewer operations, which may be performed sequentially or in parallel (eg, using parallel processors or a multi-threaded environment).

[0018] See also Figure 1 This embodiment can provide a method for determining user credit. The method for determining user credit can be used to efficiently and accurately determine the user's credit based on the user's behavioral characteristics by using a target prediction model trained by combining a particle swarm algorithm, a feedforward neural network, and an Adam optimization algorithm. The above method for determining user credit can include the following steps.

[0019] S101: Obtaining behavior information of target users from multiple data sources.

[0020] In this embodiment, since the system may include multiple data sources, in order to ensure the integrity of the obtained behavior information of the target user, the behavior information of the target user can be obtained from multiple data sources. For example, a bank system may include 5 data sources: customer liability information table, overdue repayment information table, high court dishonest customer table, customer asset table, customer personal information table, and the behavior information of the target user can be obtained from these data sources respectively. Of course, the method of obtaining the behavior information of the target user is not limited to the above examples. The technical personnel in the relevant field may make other changes under the inspiration of the technical essence of the embodiments of this specification, but as long as the functions and effects achieved are the same or similar to the embodiments of this specification, they should be covered within the protection scope of the embodiments of this specification.

[0021] In this embodiment, the above behavior information can be used to characterize the historical resource transfer behavior of the target user, such as: resource transfer amount, resource transfer time, etc. The specific information can be determined according to the actual application scenario, and this specification embodiment does not limit this.

[0022] In this implementation, the target user may be a user in the system whose credit rating needs to be evaluated, or may be a user outside the system. The specific target user may be determined based on actual conditions, and this specification does not limit this.

[0023] S102: Merging the behavior information of the target user in multiple data sources according to key elements to obtain a behavior feature information set of the target user.

[0024] In this embodiment, in order to reflect the complete behavior information of the target user, the behavior information of the target user in the multiple data sources can be merged according to the key elements, so as to ensure that all the information under the same key element can be aggregated together, and pieced together to obtain a complete behavior information with natural persons as the basic granularity. For example, all information under different customer numbers, different bank card numbers, and different bank accounts held by the same user can be aggregated together according to name, ID type, and ID number. Of course, the method of merging according to key elements is not limited to the above examples. Technical personnel in the relevant field may make other changes under the inspiration of the technical essence of the embodiments of this specification, but as long as the functions and effects achieved are the same or similar to the embodiments of this specification, they should be covered within the protection scope of the embodiments of this specification.

[0025] In this embodiment, since the initially obtained behavioral information may not fully meet the requirements of the target prediction model input, it can be processed when merging the behavioral information of the target user in multiple data sources, for example: accumulation, maximum value, minimum value, etc., to obtain multiple behavioral characteristics of the target user, and then obtain the behavioral characteristic information set of the target user.

[0026] In this embodiment, the above key elements may be one or more, and the key elements may be used to uniquely identify a user. The above key elements may be name, certificate type, and certificate number. Of course, it can be understood that the above key elements may also be other information, such as bank card number, telephone number, user number, etc. The specific information may be determined according to actual conditions, and the embodiments of this specification do not limit this.

[0027] In this embodiment, the target user's behavior feature information set may include behavior features of multiple target users, and the behavior feature information set may be stored in the form of a table, text, etc. The specific information may be determined according to actual conditions, and this embodiment of the specification does not limit this.

[0028] S103: Based on the target user's behavioral feature information set, a target prediction model is used to determine the target user's characterization value; wherein the target prediction model is a model trained by combining a particle swarm algorithm, a feedforward neural network, and an Adam optimization algorithm for determining the user's credit according to the user's behavioral features, and the characterization value is used to represent the target user's credit.

[0029] In this embodiment, the characterization value of the target user can be determined based on the target user's behavioral feature information set using a target prediction model. The target prediction model can be a neural network model pre-trained by combining a particle swarm algorithm, a feedforward neural network, and an Adam optimization algorithm to determine the user's creditworthiness based on the user's behavioral characteristics. The Adam (Adaptive moment estimation) optimization algorithm is a first-order optimization algorithm that can replace the traditional stochastic gradient descent process, and it can iteratively update the neural network weights based on training data.

[0030] In this embodiment, the particle swarm algorithm (PSO, Particle Swarm Optimization) is also called the particle swarm optimization algorithm or the bird flock foraging algorithm, which is an evolutionary computing technology. Based on the observation of the activity behavior of animal clusters, the particle swarm algorithm uses the information sharing of individuals in the group to make the movement of the entire group evolve from disorder to order in the problem-solving space, thereby obtaining the optimal solution. The above-mentioned feedforward neural network can be a BP (Back Propagation) neural network, which is a multi-layer feedforward neural network trained according to the error back propagation algorithm.

[0031] In this embodiment, the input data of the target prediction model are multiple behavioral features of the target user, and the output data are the characterization values ​​of the target user. Among them, the characterization value can be used to identify the credit of the target user. In some embodiments, the characterization value can be a value greater than 0, for example: 0, 1, 0 means that the target user has a high credit and there is no credit risk, and 1 means that the target user has a low credit and there is a credit risk. Of course, the characterization value is not limited to the above examples, and can also be high, medium, low and other values. The technical personnel in the relevant field may make other changes under the inspiration of the technical essence of the embodiments of this specification, but as long as the functions and effects achieved are the same or similar to the embodiments of this specification, they should be covered within the protection scope of the embodiments of this specification.

[0032] From the above description, it can be seen that the embodiments of this specification achieve the following technical effects: the behavior information of the target user can be obtained from multiple data sources, and the behavior information of the target user in the multiple data sources can be merged according to key elements to obtain a behavior feature information set of the target user. Furthermore, based on the behavior feature information set of the target user, the characterization value of the target user can be determined by using a target prediction model obtained by combining the particle swarm algorithm, the feedforward neural network and the Adam optimization algorithm, wherein the above-mentioned characterization value is used to represent the credit of the target user. Due to the use of the target prediction model obtained by combining the particle swarm algorithm, the feedforward neural network and the Adam optimization algorithm, the credit of the user can be determined efficiently and accurately based on the user's behavioral characteristics.

[0033] In one embodiment, merging the target user's behavior information in the multiple data sources according to key elements to obtain the target user's behavior feature information set may include: obtaining a primary and secondary guest editor relationship table, and accumulating and calculating the target user's behavior information in the multiple data sources according to key elements in the primary and secondary guest editor relationship table to obtain multiple behavior features. Further, the multiple behavior features may be used as the target user's behavior feature information set.

[0034] In this embodiment, since the information recorded in the database is generally stored at the granularity of "customer number", for the same user, since he may be involved in multiple businesses, he may have multiple user numbers in the system. For example, in a banking system, the same user may have a credit card and a passbook, which correspond to two customer numbers. Therefore, all information under the name of the same natural person can be merged based on the key element information in the primary and secondary customer code relationship table.

[0035] In this embodiment, the idea of ​​the above-mentioned main and secondary guest code relationship table is: select the earliest account opening record as the main guest code, and the subsequent opening is regarded as the secondary guest code. The main and secondary guest code relationship table can contain multiple key elements, such as: name, certificate type, certificate number, main guest code, general guest code, etc. The specific can be determined according to the actual situation, and this embodiment of the specification does not limit this.

[0036] In this embodiment, since the initially obtained behavior information may be different from the behavior features required to be input by the target prediction model, for example, the obtained behavior information is the details of all historical resource transfers of the user, while the behavior features required to be input are the maximum amount of historical resource transfers, then it is necessary to perform cumulative calculations on the behavior information to obtain the corresponding behavior features. The cumulative calculation processing method may include: summing, taking the maximum value, taking the minimum value, taking the average value, etc., which can be determined according to the actual situation and is not limited in this embodiment of the specification.

[0037] In one embodiment, the behavioral characteristics may include: the number of overdue platforms within a preset time period, the total number of overdue times within a preset time period, the total amount of overdue within a preset time period, the maximum amount of overdue within a preset time period, the number of successful repayments within a preset time period, the total amount of successful repayments within a preset time period, the maximum amount of successful repayments within a preset time period, the average monthly repayable amount within a preset time period, the maximum monthly repayable amount within a preset time period, the number of application platforms within a preset time period, the total number of applications within a preset time period, the total amount of applications within a preset time period, the average monthly repayable amount for applications within a preset time period, the maximum monthly repayable amount for applications within a preset time period, etc.

[0038] In this implementation, in the application scenario of loans, the target user may be a user who is currently applying for a loan or an external user, and the behavior information may include repayment information and application information, etc. The above-mentioned preset time period may be within the three months before the current moment, within the six months before the current moment, or all moments in the historical records, which may be determined according to actual conditions, and this specification embodiment does not limit this.

[0039] In some embodiments, the behavior characteristics included in the behavior characteristic information set may be as shown in Table 1.

[0040] Table 1

[0041]

[0042]

[0043] In one embodiment, the key elements may include: name, certificate type, certificate number, etc.

[0044] In this embodiment, the above key elements may be one or more, and the key elements may be used to uniquely identify a user. Of course, it is understandable that the above key elements may also be other information, such as bank card number, telephone number, etc. The specific information may be determined according to actual conditions, and the embodiments of this specification do not limit this.

[0045] In one embodiment, before determining the characterization value of the target user using the target prediction model based on the target user's behavior feature information set, the method may further include: obtaining an initial behavior feature sample information set; wherein the initial behavior feature sample information set includes multiple groups of behavior features. The initial behavior feature sample information set may be preprocessed to obtain a training sample information set. Furthermore, the training sample information set may be used in combination with a particle swarm algorithm, a feedforward neural network, and an Adam optimization algorithm for training to obtain the target prediction model.

[0046] In this embodiment, an initial behavior feature sample information set can be generated based on the user behavior information recorded in the system, wherein the initial behavior feature sample information set can include multiple groups of behavior features, and each group of behavior features corresponds to a sample user. In some embodiments, the above-mentioned initial behavior feature sample information set can also include labels (characterization values) of each sample user. For example, 0-1 labels can be added to the currently known classifications of dishonest customers and customers with good credit for model training in subsequent stages, and the label value is set to 1 for users with records in the overdue repayment-related data table, and vice versa. 0. Of course, the way of setting the characterization value is not limited to the above examples. Under the inspiration of the technical essence of the embodiments of this specification, technical personnel in the relevant field may also make other changes, but as long as the functions and effects achieved are the same or similar to the embodiments of this specification, they should be covered within the protection scope of the embodiments of this specification.

[0047] In this embodiment, each set of data in the above-mentioned initial behavior feature sample information set may also include the primary customer number of the corresponding sample user, which is used to identify different sample users. Of course, it can be understood that other information may also be used to identify sample users, which may be determined according to actual conditions, and this embodiment of the specification does not limit this.

[0048] In this embodiment, batch collection and calculation of new data can be performed on the Hadoop platform at a fixed frequency, such as daily or monthly, and included in the incremental table. Then, the various data indicators maintained in the full table are updated to form the latest data available for training, thereby ensuring the validity and timeliness of the data. Among them, Hadoop is a distributed system infrastructure, and users can develop distributed programs without understanding the underlying details of the distribution. Make full use of the power of the cluster for high-speed computing and storage.

[0049] In this embodiment, since the format of the initial behavior feature sample information set may not meet the input requirements of the model, the initial behavior feature sample information set may be preprocessed to obtain a training sample information set. The training sample information set is then used in combination with a particle swarm algorithm, a feedforward neural network, and an Adam optimization algorithm to perform training to obtain the target prediction model.

[0050] In this embodiment, in order to obtain a globally optimal and fast feedforward network, a method combining the PSO algorithm and the BP algorithm can be used for training: first, the PSO algorithm and the characteristics of global optimization and fast convergence speed are used to train the weights of the network. In the space close to the optimal solution obtained by using the BP algorithm, the BP algorithm is used to further optimize the optimal value of the network weights. Assuming that the user's behavioral characteristics are extracted into a data of 20 fields, the number of input layer nodes for the BP neural network is 20. Assuming that the model has a middle layer of 10 nodes and an output layer of 2 nodes (indicating the output of users with high credit and users with low credit), plus the number of bias values ​​of each layer of nodes, there are a total of 20×10+10+10×2+2=232 parameters for the entire model. Such a set of complete parameters can be regarded as a particle, and a speed attribute of the same dimension is given to it. Several such particles can be randomly generated at the beginning of the training model, so that they gradually converge to the global optimal solution during the iteration process.

[0051] In this embodiment, a complete PSO-BP can be regarded as a weak classifier, and several PSO-BPs can be integrated into a strong classifier using the Adam optimization algorithm, so as to train and obtain the target prediction model.

[0052] In one embodiment, preprocessing the behavior feature sample information set to obtain a training sample information set may include: using a synthetic minority class oversampling technique to perform equalization processing on the initial behavior feature sample information set to obtain a first behavior feature sample information set. The first behavior feature sample information set may be normalized to obtain a second behavior feature information set. Furthermore, a training sample information set and a test sample information set may be randomly generated based on the second behavior feature information set.

[0053] In this embodiment, since the data volume distribution of positive samples (high credit) and negative samples (low credit) in the actual collected behavioral feature sample data is unbalanced, it is necessary to use the synthetic minority class oversampling technology (SMOTE) to balance the initial behavioral feature sample information set to obtain the first behavioral feature sample information set. For example, in a loan scenario, loan users account for a very small part of the total users, and overdue and untrustworthy users account for a very small part of loan users, that is, the number of overdue and untrustworthy users is far less than the number of normal users. If the original data is directly used for training without preprocessing, it will cause the model to be seriously biased and unable to correctly judge potential risks. Therefore, in the data preprocessing stage, the synthetic minority class oversampling technology can be adopted to artificially synthesize the samples of the smaller number based on the similarity of the sample classes in the sample space. According to the unbalanced ratio between the number of users with low credit and the number of users with high credit, the parameters can be set to synthesize the user samples with low credit.

[0054] In this embodiment, when an initial behavior feature sample information set is given:

[0055] T={(x1,y1),(x2,y2),…,(x n ,y n )}

[0056] Among them, x n is the behavioral characteristics of the nth sample user, y n is the label (characteristic value, 1 or 0) of the nth sample user. The sample generation strategy is, for each data x in the subset of users with low credit i , identify the k nearest neighbors in the sample space, and generate k sample users with low artificial credibility according to the following strategy.

[0057]

[0058] Among them, x new is a sample user with low artificial credibility; δ is a random number between (0,1]; x i is the behavioral characteristic of the i-th sample user; is the sample x i One of the k nearest neighbor samples in the sample space. For binary classification problems such as this example, the positive and negative ratios of samples can be simply controlled in this way to optimize the model training results. The process of balancing can be as follows Figure 2 As shown in , the circular part is the sample user with high credit, the triangular part is the sample user with low credit, and the square part is the sample user with artificial low credit.

[0059] In this embodiment, since the differences between different behavioral features will vary significantly according to the business scenario, the distribution is uneven, and there is a huge order of magnitude difference between the maximum and minimum values. If the original data and artificially generated data are directly used for the next stage of model training, the training results of the entire model will be seriously biased towards one or several intervals with more training samples, and other data will be relatively meaningless, resulting in data skew. Therefore, the first behavioral feature sample information set can be normalized to obtain the second behavioral feature information set, so that all behavioral features in each training sample have the same level of influence on the sample distance in the feature space.

[0060] In this embodiment, the second behavior feature information set can be randomly generated into a training sample information set and a test sample information set according to a preset ratio. The above preset ratio can be 3:2 or 4:1. The specific ratio can be determined according to actual conditions and is not limited to this in the embodiments of this specification.

[0061] In one embodiment, normalizing the first behavior characteristic sample information set to obtain the second behavior characteristic information set may include: normalizing the behavior characteristics involving the amount in the first behavior characteristic sample information set to obtain the second behavior characteristic information set.

[0062] In this embodiment, in the case of financial scenarios, the above-mentioned first behavioral feature sample information set will have some behavioral features involving amounts, such as: application amount, repayment amount, overdue amount, etc. For data with this attribute, the differences between individuals will be significantly different according to the business scenario. It is manifested in uneven distribution, and there is a huge order of magnitude difference between the maximum and minimum values. If the original data and artificially generated data are used directly for the next stage of model training, the training results of the entire model will be seriously inclined to one or several intervals with a large number of transactions and transaction amounts, while other data will be relatively meaningless. Therefore, a z-score normalization strategy can be adopted for behavioral features involving amounts. Assuming that there is a behavioral feature M involving an amount, for the first behavioral feature sample information set, M can be normalized according to the following formula.

[0063]

[0064] Among them, M is the original value; M' is the normalized value; μ is the average value of this type of behavior feature in the first behavior feature sample information set; σ is the standard deviation of this type of behavior feature in the first behavior feature sample information set. In this way, the behavior features involving the amount whose maximum and minimum values ​​are unknown can still meet the relatively stable distribution of mean 0 and variance 1, and minimize the negative impact of the data in the most frequent interval on the overall model.

[0065] In one embodiment, after determining the characterization value of the target user by using a target prediction model based on the behavior feature information set of the target user, the method may further include: feeding back the characterization value of the target user to a target processing object.

[0066] In this embodiment, the trained target prediction model can be integrated into the background batch computing server, for example, deployed on Hadoop or MPP server and batch computing is performed on users who are currently applying for loans or external users, and finally the characterization value of each target user is obtained, and the determined characterization value is fed back to the terminal of the corresponding processing object. Among them, the above-mentioned processing object can be a business person, and the above-mentioned MPP (Massively Parallel Processing) has an independent disk storage system and memory system. Business data is divided into each node according to the database model and application characteristics. Each data node is connected to each other through a dedicated network or a commercial general network, and cooperates with each other to provide database services as a whole. Non-shared database clusters have the advantages of complete scalability, high availability, high performance, excellent cost performance, resource sharing, etc.

[0067] In this implementation, the processing object can clearly recognize the value of existing users by determining the characterization value of each target user, and make full use of the massive data at hand to accurately identify users who can or cannot provide loans; externally, the characterization value can be used to evaluate external data, and personal loan products can be promoted purposefully and targeted in a point-to-point manner to expand the market and screen out potential high-quality users. This can provide strong technical support for user managers as an auxiliary means to expand the market, find potential users, and mine value from massive data while controlling and avoiding risks to maximize profits.

[0068] Based on the same inventive concept, a device for determining user credit is also provided in the embodiments of this specification, as described in the following embodiments. Since the principle of solving the problem by the device for determining user credit is similar to that of the method for determining user credit, the implementation of the device for determining user credit can refer to the implementation of the method for determining user credit, and the repeated parts will not be repeated. As used below, the term "unit" or "module" can be a combination of software and / or hardware that implements a predetermined function. Although the device described in the following embodiments is preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable. Figure 3 is a structural block diagram of a device for determining user credit in an embodiment of this specification, such as Figure 3 As shown, it may include: an acquisition module 301, a merging module 302, and a determination module 303. The structure is described below.

[0069] The acquisition module 301 may be used to acquire the behavior information of the target user from multiple data sources.

[0070] The merging module 302 may be used to merge the behavior information of the target user in the multiple data sources according to key elements to obtain a behavior feature information set of the target user.

[0071] The determination module 303 can be used to determine the characterization value of the target user based on the behavioral feature information set of the target user using a target prediction model; wherein the target prediction model is a model trained by combining a particle swarm algorithm, a feedforward neural network and an Adam optimization algorithm for determining the user's credit according to the user's behavioral characteristics, and the characterization value is used to represent the credit of the target user.

[0072] The embodiment of this specification also provides an electronic device, which can be specifically referred to in Figure 4 The schematic diagram of the structure of the electronic device composition of the method for determining the user credit provided by the embodiment of the present specification is shown, and the electronic device may specifically include an input device 41, a processor 42, and a memory 43. Among them, the input device 41 can be specifically used to input the behavior information of the target user from multiple data sources. The processor 42 can be specifically used to merge the behavior information of the target user from the multiple data sources according to key elements to obtain a behavior feature information set of the target user; based on the behavior feature information set of the target user, determine the characterization value of the target user using a target prediction model; wherein the target prediction model is a model obtained by combining particle swarm algorithm, feedforward neural network and Adam optimization algorithm for training to determine the user's credit according to the user's behavior characteristics, and the characterization value is used to represent the credit of the target user. The memory 43 can be specifically used to store parameters such as the characterization value of the target user.

[0073] In this embodiment, the input device may specifically be one of the main devices for information exchange between the user and the computer system. The input device may include a keyboard, a mouse, a camera, a scanner, a light pen, a handwriting input board, a voice input device, etc.; the input device is used to input the original data and the program for processing these numbers into the computer. The input device can also obtain and receive data transmitted from other modules, units, and devices. The processor can be implemented in any appropriate manner. For example, the processor can take the form of a computer-readable medium, a logic gate, a switch, an application-specific integrated circuit (Application Specific Integrated Circuit, ASIC), a programmable logic controller, and an embedded microcontroller, etc., such as a microprocessor or a processor and a computer-readable program code (such as software or firmware) that can be executed by the (micro) processor. The memory may specifically be a memory device used to store information in modern information technology. The memory may include multiple levels. In a digital system, anything that can store binary data can be a memory; in an integrated circuit, a circuit with a storage function without a physical form is also called a memory, such as a RAM, a FIFO, etc.; in a system, a storage device with a physical form is also called a memory, such as a memory stick, a TF card, etc.

[0074] In this embodiment, the functions and effects specifically realized by the electronic device can be explained in comparison with other embodiments and will not be described in detail here.

[0075] In the implementation manner of the embodiment of this specification, a computer storage medium based on a method for determining user credit is also provided, and the computer storage medium stores computer program instructions, which can achieve the following when executed: obtaining behavior information of a target user from multiple data sources; merging the behavior information of the target user in the multiple data sources according to key elements to obtain a behavior feature information set of the target user; based on the behavior feature information set of the target user, determining a characterization value of the target user using a target prediction model; wherein the target prediction model is a model obtained by training in combination with a particle swarm algorithm, a feedforward neural network and an Adam optimization algorithm for determining a user's credit according to the user's behavior characteristics, and the characterization value is used to represent the credit of the target user.

[0076] In this embodiment, the storage medium includes, but is not limited to, a random access memory (RAM), a read-only memory (ROM), a cache, a hard disk (HDD), or a memory card. The memory may be used to store computer program instructions. The network communication unit may be an interface for network connection communication set in accordance with the standard specified by the communication protocol.

[0077] In this embodiment, the functions and effects specifically implemented by the program instructions stored in the computer storage medium can be explained in comparison with other embodiments and will not be repeated here.

[0078] Obviously, those skilled in the art should understand that the modules or steps of the above-mentioned embodiments of this specification can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, and optionally, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, and in some cases, the steps shown or described can be executed in a different order from that here, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. In this way, the embodiments of this specification are not limited to any specific combination of hardware and software.

[0079] Although the embodiments of this specification provide method operation steps as described in the above embodiments or flowcharts, more or fewer operation steps may be included in the method based on routine or no creative labor. In the steps where there is no necessary causal relationship logically, the execution order of these steps is not limited to the execution order provided by the embodiments of this specification. When the device or terminal product of the method described in practice is executed, it can be executed in sequence or in parallel according to the method shown in the embodiments or drawings (for example, a parallel processor or a multi-threaded processing environment).

[0080] It should be understood that the above description is for illustration and not for limitation. By reading the above description, many embodiments and many applications beyond the examples provided will be apparent to those skilled in the art. Therefore, the scope of the embodiments of this specification should not be determined with reference to the above description, but should be determined with reference to the full scope of the aforementioned claims and the equivalents possessed by these claims.

[0081] The above is only a preferred embodiment of the embodiment of this specification, and is not intended to limit the embodiment of this specification. For those skilled in the art, the embodiment of this specification may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the embodiment of this specification shall be included in the protection scope of the embodiment of this specification.

Claims

1. A method for determining user credit, characterized in that: include: Obtaining the target user's behavior information from multiple data sources; wherein the multiple data sources include: customer debt information table, overdue repayment information table, high court dishonest customer table, customer asset table, customer personal information table; Merging the behavior information of the target user in the multiple data sources according to key elements to obtain a behavior feature information set of the target user; Based on the behavior feature information set of the target user, a characterization value of the target user is determined using a target prediction model; wherein the target prediction model is a model obtained by training a training sample information set in combination with a particle swarm algorithm, a feedforward neural network, and an Adam optimization algorithm for determining the user's credit according to the user's behavior features, the characterization value is used to represent the credit of the target user, the training sample information set is randomly generated according to a second behavior feature information set, the second behavior feature information set is obtained by normalizing the first behavior feature sample information set, the first behavior feature sample information set is obtained by balancing the initial behavior feature sample information set using a synthetic minority class oversampling technique, and the synthetic minority class oversampling technique is used to synthesize samples corresponding to sample classes whose number of samples is less than a number threshold based on the similarity of sample classes in the sample space; Wherein, the training sample information set is constructed by incremental data obtained from a distributed system based on a preset frequency; The target prediction model is deployed on a massively parallel processing server, and the behavior information is determined based on business data stored in nodes corresponding to different businesses in the massively parallel processing server; the nodes have independent disk storage systems and memory systems.

2. The method according to claim 1, characterized in that Merging the behavior information of the target user in the multiple data sources according to key elements to obtain a behavior feature information set of the target user includes: Get the relationship table between main editor and deputy editor; According to the key elements in the main-sub-guest editor relationship table, the behavior information of the target user in the multiple data sources is accumulated and calculated to obtain multiple behavior features; The multiple behavior characteristics are used as a behavior characteristic information set of the target user.

3. The method according to claim 2, characterized in that The behavioral characteristics include: the number of overdue platforms within a preset time period, the total number of overdue times within a preset time period, the total amount of overdues within a preset time period, the maximum amount of overdues within a preset time period, the number of successful repayments within a preset time period, the total amount of successful repayments within a preset time period, the maximum amount of successful repayments within a preset time period, the average monthly repayable amount within a preset time period, the maximum monthly repayable amount within a preset time period, the number of application platforms within a preset time period, the total number of applications within a preset time period, the total amount of applications within a preset time period, the average monthly repayable amount for applications within a preset time period, and the maximum monthly repayable amount for applications within a preset time period.

4. The method according to claim 1, characterized in that: The key elements include: name, ID type, and ID number.

5. The method according to claim 1, characterized in that Before determining the representation value of the target user by using a target prediction model based on the behavior feature information set of the target user, the method further includes: Acquire an initial behavior feature sample information set; wherein the initial behavior feature sample information set includes multiple groups of behavior features; Preprocessing the initial behavior feature sample information set to obtain a training sample information set; The target prediction model is obtained by training using the training sample information set in combination with a particle swarm algorithm, a feedforward neural network and an Adam optimization algorithm.

6. The method according to claim 5, characterized in that Preprocessing the behavior feature sample information set to obtain a training sample information set includes: Using a synthetic minority class oversampling technique to perform equalization processing on the initial behavior feature sample information set to obtain a first behavior feature sample information set; Normalizing the first behavior feature sample information set to obtain a second behavior feature information set; A training sample information set and a test sample information set are randomly generated according to the second behavior feature information set.

7. The method according to claim 6, characterized in that Normalizing the first behavior characteristic sample information set to obtain a second behavior characteristic information set includes: normalizing the behavior characteristics involving the amount in the first behavior characteristic sample information set to obtain the second behavior characteristic information set.

8. The method according to claim 1, characterized in that After determining the characterization value of the target user by using a target prediction model based on the behavior feature information set of the target user, the method further includes: The characterization value of the target user is fed back to the target processing object.

9. A device for determining user credit, characterized in that: include: The acquisition module is used to obtain the behavior information of the target user from multiple data sources; A merging module, configured to merge the behavior information of the target user in the multiple data sources according to key elements to obtain a behavior feature information set of the target user; A determination module, configured to determine the characterization value of the target user based on the target user's behavior feature information set using a target prediction model; wherein the target prediction model is a model trained by combining a particle swarm algorithm, a feedforward neural network, and an Adam optimization algorithm to determine the user's creditworthiness according to the user's behavior features, and the characterization value is used to represent the target user's creditworthiness; Wherein, the training sample information set is constructed by incremental data obtained from a distributed system based on a preset frequency; The target prediction model is deployed on a massively parallel processing server, and the behavior information is determined based on business data stored in nodes corresponding to different businesses in the massively parallel processing server; the nodes have independent disk storage systems and memory systems.

10. A device for determining user credit, characterized in that: The method comprises a processor and a memory for storing processor-executable instructions, wherein the processor implements the steps of the method according to any one of claims 1 to 8 when executing the instructions.

11. A computer-readable storage medium, characterized in that: Computer instructions are stored thereon, and when the instructions are executed, the steps of the method according to any one of claims 1 to 8 are implemented.

Images