Credit card risk category determination method and device, equipment, medium and program product

Abstract

The embodiment of the invention provides a credit card risk category determination method and device, equipment, a medium and a program product. The method comprises the steps that a sample set is obtained, the sample set comprises data information of n credit cards, the n credit cards are marked with respective risk categories, the data information of the n credit cards corresponds to m data categories, any credit card comprises data information under the m data categories, and m and n are integers larger than 1; determining whether each data category in the sample set is associated with the risk category through a fuzzy rough set model, and determining the data category associated with the risk category as a target data category, the fuzzy rough set model being a model for determining the association according to a fuzzy approximation operator; and under the condition that the at least one target data category is determined, inputting the data information of the credit card to be tested under the at least one target data category into a classification model for classification prediction, and determining the risk category of the credit card to be tested. The method can improve the determination efficiency.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a method, device, equipment, medium and program product for determining a credit card risk category. Background Art

[0002] With the development of credit card business, the number of users using credit cards is increasing. In order to better meet the needs of users, the use scenarios of credit cards are also diversified. For example, credit cards can be used in online payment, offline payment or online transfer and other use scenarios.

[0003] Some users may violate the rules when using credit cards, such as forging credit cards, impersonating others, or maliciously overdrawing. Therefore, it is necessary for credit card issuers to determine whether there are card usage risks during the use of credit cards, and to promptly identify credit cards that may be risky, so as to facilitate card management.

[0004] In the related art, when determining whether a credit card is risky, the collected data information of the credit card user is input into a pre-trained prediction model for processing to obtain a prediction result that characterizes whether the credit card is risky. When using this method for prediction, the amount of calculation required is large and the time consumption is large, resulting in low efficiency in determining the risk category. If the prediction model is changed to reduce the amount of calculation in order to improve efficiency, the accuracy of the prediction result may be reduced. Therefore, the prior art has the problem of not being able to improve the determination efficiency while taking into account the prediction accuracy. Summary of the invention

[0005] The embodiments of the present application provide a credit card risk category determination method, apparatus, device, medium and program product, so as to achieve the effect of improving the determination efficiency while taking into account the accuracy of determining the credit card risk category.

[0006] In a first aspect, an embodiment of the present application provides a method for determining a credit card risk category, comprising: obtaining a sample set, the sample set comprising data information of n credit cards, the n credit cards being each marked with their own risk category, the data information of the n credit cards corresponding to m data categories, any one of the credit cards comprising data information under the m data categories, and both m and n being integers greater than 1; determining whether each data category in the sample set is associated with a risk category through a fuzzy rough set model, and determining the data category associated with the risk category as a target data category, the fuzzy rough set model being a model for determining association based on a fuzzy approximate operator; when at least one target data category is determined, inputting the data information of the credit card to be tested under the at least one target data category into a classification model for classification prediction, and determining the risk category of the credit card to be tested.

[0007] In a possible implementation, the method of determining whether each data category in the sample set is associated with a risk category through a fuzzy rough set model, and determining the data category associated with the risk category as a target data category, includes: selecting one or more data categories from each data category in the sample set to obtain a first category subset; adding one or more unselected data categories to the first category subset, or removing one or more data categories from the first category subset to obtain a second category subset; calculating the fuzzy dependency of the data information of each credit card in the sample set under the first category subset on the risk category through the fuzzy rough set model to obtain a first fuzzy dependency, and calculating the fuzzy dependency of the data information of each credit card in the sample set under the second category subset on the risk category to obtain a second fuzzy dependency; determining whether the data category added or removed when obtaining the second category subset is associated with the risk category based on the first fuzzy dependency and the second fuzzy dependency, and determining the data category associated with the risk category as the target data category.

[0008] In a possible implementation, the following formula is satisfied when the fuzzy dependency is calculated by the fuzzy rough set model: Wherein, U represents the sample set; x i represents the data information of the i-th credit card in the sample set; L represents each risk category; M represents the distance measurement matrix; B represents the category subset corresponding to the calculated fuzzy dependency; Represents x i ∈U is the fuzzy lower approximation operator based on B for each risk category; Indicates fuzzy positive area; |*| indicates modulus; Indicates fuzzy dependency.

[0009] In a possible implementation, the distance metric matrix is ​​determined after iterative learning through a large-margin nearest neighbor algorithm based on data information in the sample set.

[0010] In a possible implementation, the distance metric matrix is ​​a Markov matrix, and the Markov matrix satisfies the following formula: Wherein, M represents the Markov matrix; x j represents the data information of the jth credit card in the sample set; D M (x i ,x j ) represents x i With x j Markov distance; when the large margin nearest neighbor algorithm is used for iterative learning, the loss function used satisfies the following formula: ε(M)=(1-μ)εpull (M)+με push (M); Among them, ε(M) represents the value of the loss function; μ represents the preset coefficient; SKNN(x i ) indicates that x i The set of k nearest neighbors with the same risk category, k is any positive integer; x l Indicates that it belongs to KNN(x i ) set and does not belong to SKNN(x i ) data information of the lth credit card in the set, KNN(x i ) is the same as x i The set of the nearest k neighbors, k is any positive integer; [Z] + =max(Z,0) means the value in the brackets is compared with 0 and the larger one is taken; y il Represents x i With x j The corresponding values ​​for the same or different risk categories, x i With x j When the risk categories are the same, the value is 1, x i With x j The value is 0 when the risk categories are different.

[0011] In a possible implementation, the step of inputting data information of the credit card to be tested under the at least one target data category into a classification model for classification prediction to determine the risk category of the credit card to be tested includes: inputting data information of the credit card to be tested under the at least one target data category into the classification model to calculate the function value of the credit card to be tested corresponding to the objective function; performing classification prediction based on the function value and a preset judgment threshold to determine the risk category of the credit card to be tested; wherein the objective function is used to calculate the probability of the risk category corresponding to the credit card to be tested, the objective function is preset based on the support vector distance between the risk category of the credit card in the sample set and the high-dimensional discriminant function, the high-dimensional discriminant function is obtained by transforming the linear discriminant function through the Gaussian kernel, and the linear discriminant function is determined under the at least one target data category based on the data information in the sample set using a support vector machine model.

[0012] In a second aspect, an embodiment of the present application provides a device for determining a credit card risk category, comprising: an acquisition module, used to acquire a sample set, the sample set comprising data information of n credit cards, the n credit cards being marked with their respective risk categories, the data information of the n credit cards corresponding to m data categories, any one of the credit cards comprising data information under the m data categories, wherein m and n are both integers greater than 1; a determination module, used to determine whether each data category in the sample set is associated with a risk category through a fuzzy rough set model, and to determine the data category associated with the risk category as a target data category, the fuzzy rough set model being a model for determining the association based on a fuzzy approximate operator; a processing module, used to, when at least one target data category is determined, input the data information of the credit card to be tested under the at least one target data category into a classification model for classification prediction, so as to determine the risk category of the credit card to be tested.

[0013] In one possible implementation, the determination module is specifically used to: select one or more data categories from each data category in the sample set to obtain a first category subset; add one or more unselected data categories to the first category subset, or remove one or more data categories from the first category subset to obtain a second category subset; calculate the fuzzy dependency of the data information of each credit card in the sample set in the first category subset on the risk category through a fuzzy rough set model to obtain a first fuzzy dependency, and calculate the fuzzy dependency of the data information of each credit card in the sample set in the second category subset on the risk category to obtain a second fuzzy dependency; determine whether the data categories added or removed when obtaining the second category subset are associated with the risk category based on the first fuzzy dependency and the second fuzzy dependency, and determine the data categories associated with the risk category as target data categories.

[0014] In a possible implementation, the following formula is satisfied when the fuzzy dependency is calculated by the fuzzy rough set model: Wherein, U represents the sample set; x i represents the data information of the i-th credit card in the sample set; L represents each risk category; M represents the distance measurement matrix; B represents the category subset corresponding to the calculated fuzzy dependency; Represents x i ∈U is the fuzzy lower approximation operator based on B for each risk category; Indicates fuzzy positive area; |*| indicates modulus; Indicates fuzzy dependency.

[0015] In a possible implementation, the distance metric matrix is ​​determined after iterative learning through a large-margin nearest neighbor algorithm based on data information in the sample set.

[0016] In a possible implementation manner, the distance metric matrix is ​​a Markov matrix, and the Markov matrix satisfies the following formula: M (x i ,x j )=(x i -x j ) T M(x i -x j ), wherein M represents the Markov matrix; x j represents the data information of the jth credit card in the sample set; D M (x i ,x j ) represents x i With x j Markov distance; when the large margin nearest neighbor algorithm is used for iterative learning, the loss function used satisfies the following formula: ε(M)=(1-μ)ε pull (M)+με push (M); Among them, ε(M) represents the value of the loss function; μ represents the preset coefficient; SKNN(x i ) indicates that x i The set of k nearest neighbors with the same risk category, k is any positive integer; x l Indicates that it belongs to KNN(x i ) set and does not belong to SKNN(x i ) data information of the lth credit card in the set, KNN(x i ) is the same as x i The set of the nearest k neighbors, k is any positive integer; [Z] + =max(Z,0) means the value in the brackets is compared with 0 and the larger one is taken; y il Represents x i With x j The corresponding values ​​for the same or different risk categories, x i With x j When the risk categories are the same, the value is 1, x i With x j The value is 0 when the risk categories are different.

[0017] In one possible implementation, the processing module is specifically used to: input the data information of the credit card to be tested under the at least one target data category into the classification model to calculate the function value of the credit card to be tested corresponding to the objective function; perform classification prediction based on the function value and a preset judgment threshold to determine the risk category of the credit card to be tested; wherein the objective function is used to calculate the probability of the risk category corresponding to the credit card to be tested, the objective function is preset based on the support vector distance between the risk category of the credit card in the sample set and the high-dimensional discriminant function, the high-dimensional discriminant function is obtained by transforming the linear discriminant function through the Gaussian kernel, and the linear discriminant function is determined under the at least one target data category using the support vector machine model based on the data information in the sample set.

[0018] In a third aspect, an embodiment of the present application provides an electronic device, comprising: a memory and a processor; the memory stores computer-executable instructions; the processor executes the computer-executable instructions stored in the memory, so that the processor executes the first aspect above and / or various possible implementations of the first aspect.

[0019] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are executed by a processor, they are used to implement the first aspect above and / or various possible implementations of the first aspect.

[0020] In a fifth aspect, an embodiment of the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the above first aspect and / or various possible implementation methods of the first aspect.

[0021] The credit card risk category determination method, device, equipment, medium and program product provided by the embodiment of the present application obtain a sample set through the data information of the credit card marked with the risk category, and determine which data categories in the sample set are associated with the risk category through the fuzzy rough set model, and determine the target data category associated with the risk category. Since the fuzzy rough set model can determine the association according to the fuzzy approximate operator, it is possible to more accurately screen out the data category that is helpful in determining the risk category from the numerous data categories, thereby achieving the purpose of filtering out the data categories that are irrelevant to the determination of the risk category, reducing the number of data categories, and further reducing the amount of data information. When at least one target data category is determined, the data information of the credit card to be tested under at least one target data category is input into the classification model for classification prediction. Compared with the case where the data information under all data categories of the credit card to be tested is input, this method can reduce the amount of data input into the classification model, reduce the amount of calculation when the classification model is calculated, and further shorten the time required for calculation, thereby improving the efficiency of classification prediction. Since the filtered data categories are irrelevant data categories, the interference of irrelevant data information is eliminated, so that the accuracy of the prediction result is higher. Therefore, based on this method, the effect of improving the determination efficiency can be achieved while taking into account the accuracy of determining the credit card risk category. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0023] Figure 1 A schematic diagram of an application scenario provided in an embodiment of the present application;

[0024] Figure 2 Schematic diagram of the process of determining the credit card risk category provided in the embodiment of the present application Figure 1 ;

[0025] Figure 3 Schematic diagram of the process of determining the credit card risk category provided in the embodiment of the present application Figure 2 ;

[0026] Figure 4 Schematic diagram of the process of determining the credit card risk category provided in the embodiment of the present application Figure 3 ;

[0027] Figure 5 A schematic diagram of the structure of a device for determining a credit card risk category provided in an embodiment of the present application;

[0028] Figure 6 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application.

[0029] The above drawings have shown clear embodiments of the present application, which will be described in more detail later. These drawings and text descriptions are not intended to limit the scope of the present application in any way, but to illustrate the concept of the present application to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0030] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application. Instead, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the appended claims.

[0031] In the embodiments of the present application, words such as "first" and "second" are used to distinguish the same or similar items with substantially the same functions and effects. For example, the first electronic device and the second electronic device are only used to distinguish different electronic devices, and their order is not limited. Those skilled in the art can understand that words such as "first" and "second" do not limit the quantity and execution order, and words such as "first" and "second" do not necessarily limit them to be different.

[0032] In the embodiments of the present application, the words "exemplary" or "for example" are used to indicate examples, illustrations or descriptions. Any embodiment or design described as "exemplary" or "for example" in the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary" or "for example" is intended to present related concepts in a specific way.

[0033] In the embodiments of the present application, "at least one" means one or more, and "more" means two or more. "And / or" describes the association relationship of the associated objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone, where A and B can be singular or plural. The character " / " generally indicates that the associated objects before and after are in an "or" relationship.

[0034] In the technical solutions of the embodiments of the present application, the collection, storage, use, processing, transmission, provision and disclosure of user personal information involved are in compliance with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0035] 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 the relevant laws, regulations and standards of the relevant countries and regions, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0036] With the rapid development of the credit card industry, banks and other credit card issuers have been continuously launching new credit card products and expanding card issuance channels in order to expand their market share. They have also actively cooperated with Internet companies and used big data and Internet technologies to deeply explore user needs and consumption patterns. In terms of product innovation, credit card services are becoming more personalized and distinctive to meet the needs of different consumer groups. At the same time, card issuers are also improving service quality and focusing on humanized services to enhance user stickiness and satisfaction.

[0037] With the popularity of mobile payment, the use of credit cards has become more diversified, and consumers can make online and offline payments more conveniently. However, with the growth of credit card business, problems such as credit card fraud and other illegal use of credit cards have become increasingly prominent. Credit card fraud includes forging credit cards, impersonating others' credit cards, and malicious overdrafts, which will bring risks and losses to banks and consumers. In order to meet these challenges, banks and other financial institutions are taking a variety of measures to strengthen risk management, improve credit card security, and promote financial knowledge education.

[0038] In this process, banks and credit card companies are also using financial technology such as machine learning algorithms to predict and prevent credit card fraud and reduce losses. The use of these technologies not only improves the security of credit card business, but also provides banks with more accurate risk assessment and decision-making support.

[0039] Figure 1 A schematic diagram of an application scenario provided in an embodiment of the present application, such as Figure 1 As shown in the figure, among many credit cards, there may be risky credit cards that may be used for credit card fraud. The credit card manager can collect the data information of multiple credit cards to be tested and build a data table including multiple credit card data information. The data table is processed by machine algorithms such as prediction models to identify risky credit cards and obtain the prediction results of the risk categories of each credit card. Then, corresponding management measures can be taken for risky credit cards to better manage card usage behavior.

[0040] In the related art, there is a method for predicting the risk of credit card fraud. The method first collects cardholder data information, which includes basic data information, financial operation behavior data and credit card transaction data information. The cardholder's basic data information, financial operation behavior data and credit card transaction data information are embedded and represented by a feedforward neural network model, a convolutional neural network model and a bidirectional long short-term memory network model (Bi-directional longshort-term memory, Bi-LSTM), and the sample feature representation vector is obtained by splicing. A generative adversarial network model is constructed and trained, and simulated fraud samples are generated using the trained generator model, and a new training set is generated by merging. Multiple detection base models are constructed and trained, and the multiple trained detection base models are combined with the discriminator model to obtain a credit card fraud detection integrated model. When a new credit card transaction occurs, the credit card risk category is determined as having a fraud risk or not having a fraud risk by the credit card fraud detection integrated model.

[0041] Using feedforward neural network models, convolutional neural network models and Bi-LSTM models for feature embedding representation, combined with a generative adversarial network model, such a model structure is relatively complex and may lead to low computational efficiency, especially when processing large-scale data sets, the demand for computing resources may become a limiting factor.

[0042] It can be seen that in this method, the model used to determine the risk of credit card fraud is obtained by combining multiple relatively complex models. Such a model is highly complex and usually includes a complex model structure and a large number of model parameters. When performing reasoning operations, the model needs to consume a large amount of computing resources and the time required for the operation is usually high. In addition, the collected data information includes basic data information, financial operation behavior data, and credit card transaction data information. These data information correspond to multiple data categories, which can be understood as high-dimensional data information. If these high-dimensional data information are not effectively deleted, the amount of data input to the model will be large, which will also increase the operation time and affect efficiency. Furthermore, with the increase in the number of credit card users, the number of credit cards that need to be risk-detected will increase. If a large amount of credit card data information is processed in batches, the efficiency of credit card risk determination will be further reduced.

[0043] The inventors have found through research that if conventional improvement ideas are followed, the structure or parameters of the model will usually be improved to simplify the model and reduce the complexity of the model. In this way, the calculation steps or calculation difficulty of the model during reasoning can be reduced, the calculation time consumption can be reduced, and the efficiency of risk classification determination can be improved.

[0044] However, if the structure or parameters of the model are simplified, the accuracy of data processing may be affected. For example, after reducing the number of convolutional layers in the convolutional neural network model, although the amount of convolution operations can be reduced, the representation strength of the output data of the convolutional neural network model may decrease, which may ultimately reduce the accuracy of the prediction results. For example, a credit card with a risk of fraud may be predicted as a credit card without a risk of fraud, or a credit card without a risk of fraud may be predicted as a credit card with a risk of fraud.

[0045] In view of this, in order to improve the efficiency of determination while taking into account the accuracy of credit card risk classification, the embodiment of the present application provides a credit card risk classification determination method, which can screen multiple data categories corresponding to the data information of the credit card based on the sample set, and determine the data category associated with the determination of the credit card risk category as the target data category, which is equivalent to reducing the data volume of the data information by reducing the dimension of the data information. In addition, the screened data category is a data category that is not associated with the risk category, which not only reduces the data volume of the data information during prediction, but also excludes irrelevant data information, which not only does not reduce the accuracy of the prediction, but also may improve the accuracy of the prediction by removing the interfering data information. In addition, based on this method, the amount of calculation during model reasoning can be reduced without changing the model structure or parameters, and the prediction efficiency of the model can be improved. Therefore, through this method, the accuracy and efficiency of credit card risk classification determination can be taken into account, and the efficiency of credit card risk classification determination can be effectively improved.

[0046] The technical solution of the present application is described in detail with specific embodiments below. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present application will be described below in conjunction with the accompanying drawings.

[0047] Figure 2 Schematic diagram of the process of determining the credit card risk category provided in the embodiment of the present application Figure 1 The execution subject of the method may be an electronic device with corresponding data storage and computing capabilities, such as a computer, a server or a server cluster. Figure 2 As shown, the method includes:

[0048] S201, obtaining a sample set, the sample set including data information of n credit cards, each of the n credit cards is marked with its own risk category, the data information of the n credit cards corresponds to m data categories, any credit card includes data information under m data categories, and m and n are both integers greater than 1.

[0049] Exemplarily, the data information of a credit card may be any data information associated with the credit card. The data information may include basic information of the credit card holder, credit card transaction information, credit card information, and financial operation behavior information of the credit card holder, etc. For example, it may include the credit card number, credit card validity period, credit card account balance, card holder asset evaluation information, security code, card issuing institution information, credit limit, billing address, credit card transaction history data, and card holder associated bank card information, etc.

[0050] Data categories can be understood as categories corresponding to specific data in data information. For example, the credit card number is a data category, and the data information of the first credit card among n credit cards under this data category is the specific card number of the credit card, such as "XXXX XXXX XXXX". For another example, the credit card account balance is a data category, and the data information of the first credit card among n credit cards under this data category is the specific balance value of the credit card, such as "200.01".

[0051] It should be understood that data categories can also be understood as the dimensions of data information. For example, if m is 100, that is, there are 100 data categories in the sample set, then the dimension of the data information is 100, and the data information of any credit card can be represented as a 100-dimensional vector. Assuming that the first dimension is the credit card number, the second dimension is the credit card account balance, etc., the credit card data information can be represented by a multidimensional vector as (XXXX XXXX XXXX, 200.01, ...).

[0052] For ease of description and understanding, the data information corresponding to any credit card in the sample set can be understood as a sample, and the risk category of the sample can be understood as the label of the sample. When processing data information, the sample can be represented by a multidimensional vector. For example, if a sample is (XXXX XXXX XXXX, 200.01, ...), then when processing the sample, the multidimensional vector can be processed.

[0053] The data information of n credit cards is obtained by data collection. For example, the data information of multiple credit cards is obtained from a database that stores the data information of multiple credit cards by data query. When obtaining the data information, the risk category that represents the corresponding credit card as risky or risk-free can also be obtained. For example, 0 can be used to represent risk-free and 1 can be used to represent risky; or -1 can be used to represent risk-free and 1 can be used to represent risky. The risk category can also be three risk categories of low risk, medium risk and high risk, or other risk categories. The method of the embodiment of the present application can be understood as a binary classification task of two classification categories, or it can be a multi-classification task of multiple classification categories.

[0054] Alternatively, the n credit cards obtained can be marked with risk categories by manual marking and / or machine marking to obtain the risk categories corresponding to each credit card. For example, after obtaining the data information of 1,000 credit cards, the risk categories of the 1,000 credit cards are predicted by the existing classification model to obtain their respective risk categories, and then each risk category is verified and adjusted by manual verification, thereby obtaining 1,000 credit cards marked with risk categories and their data information, and then obtaining a sample set. Among them, the classification model can be a machine model obtained through training, for example, it can be a model for classification prediction obtained after supervised training of the initial deep learning model.

[0055] S202, determining whether each data category in the sample set is associated with the risk category through a fuzzy rough set model, and determining the data category associated with the risk category as the target data category. The fuzzy rough set model is a model that determines the association based on a fuzzy approximation operator.

[0056] Exemplarily, the fuzzy rough set model can be a mathematical calculation model that determines the correlation according to the fuzzy approximation operator, and the fuzzy approximation operator in the fuzzy rough set model can be a fuzzy lower approximation operator or a fuzzy upper approximation operator. The fuzzy rough set model can be constructed based on a preset distance measurement method.

[0057] For example, if the distance measurement method for calculating the distance between samples is preset to be Euclidean distance, a fuzzy equivalence relationship can be predefined under the distance measurement of Euclidean distance, and a fuzzy lower approximation operator or a fuzzy upper approximation operator can be defined based on the fuzzy equivalence relationship. Taking the fuzzy lower approximation operator as an example, a fuzzy positive region can be defined based on the fuzzy lower approximation operator. The fuzzy lower approximation operator and the fuzzy positive region are constructed as a fuzzy rough set model. Based on the fuzzy positive region in the fuzzy rough set model, the data information of multiple credit cards in the sample set can be calculated, and according to the calculated value, it can be determined whether each data category corresponding to each data information is associated with the risk category.

[0058] Among them, the data category associated with the risk category can be understood as the data category that can affect the risk category corresponding to the credit card, and the data category can be determined as the target data category. For example, after calculation through the fuzzy rough set model, if the transaction amount of the credit card can affect the risk category of the credit card, the transaction amount of the credit card can be determined as a target data category.

[0059] For example, a fuzzy rough sets based on metric learning (ML-FRS) model can be constructed. Since the distance metric in the fuzzy rough sets model is obtained based on metric learning, the distance metric can adjust the metric parameters according to the distribution characteristics of each sample in the sample set during learning. Therefore, the distance metric obtained after learning can more accurately measure the distance between any two samples, thereby improving the ability of the fuzzy rough sets model to determine the association, so that the determined target data category is a data category with a greater probability of being associated with the determined risk category, and the reliability of the target data category can be improved.

[0060] In a possible implementation, a distance metric matrix can be determined after iterative learning through a large margin nearest neighbor (LMNN) algorithm according to the data information in the sample set. Based on the obtained distance metric matrix, a fuzzy rough set model based on metric learning can be constructed.

[0061] Exemplarily, in the sample set, the data information of any credit card can be a sample for learning the distance metric matrix. The LMNN algorithm establishes a more appropriate boundary by minimizing the distance between the sample and its nearest neighbor sample, while maximizing the distance between samples with different labels from the sample, and learns a distance metric matrix. Among them, the label is the risk category corresponding to the sample. For any sample, the nearest neighbor sample to the sample can be understood as the target sample, the label of the target sample is the same as the label of any sample, and the distance is the shortest. The shortest distance can be the minimum distance calculated under any distance metric such as Euclidean distance, cosine similarity or Mahalanobis distance.

[0062] For example, when determining the risk category of a credit card, the collected data categories are usually large, and the data information of a credit card corresponds to multiple data categories. It can be understood that the vector dimension corresponding to the sample is high, and the number of samples of different risk categories is relatively unbalanced. For example, the number of credit cards marked as risky and that of credit cards marked as risk-free is quite different. Therefore, the sample set of credit cards has the characteristic of uneven distribution of samples.

[0063] The LMNN algorithm mainly learns a linear transformation to maximize the probability that the k nearest neighbors of a data point belong to the same category in the transformed space, and at the same time maximize the distance between data points of different categories. Since the Mahalanobis distance can be represented by a linear transformation and can capture the covariance structure of the data, the Mahalanobis distance is usually used as a metric in LMNN. However, the Mahalanobis distance is not suitable for the distribution characteristics of samples in the credit card sample set.

[0064] For unevenly distributed or sparse sample sets, the Markov distance metric can better handle sparse data information by focusing on state transitions rather than the frequency of occurrence of a single state. Therefore, a Markov matrix can be learned through the LMNN algorithm. The Markov matrix obtained after learning can adapt to the distribution characteristics of samples in the credit card sample set, and then more accurately measure the distance between any two samples, which helps to improve the accuracy of model calculation when using the fuzzy rough set model under this sample set.

[0065] In a possible implementation, the distance metric matrix is ​​a Markov matrix, and the Markov matrix satisfies the following formula (1):

[0066]

[0067] Where M represents the Markov matrix; x j represents the data information of the jth credit card in the sample set; D M (x i ,x j ) represents x i With x j The Markov distance of

[0068] When iterative learning is performed using the large-margin nearest neighbor algorithm, the loss function used satisfies the following formulas (2) to (4):

[0069] ε(M)=(1-μ)ε pull (M)+με push (M) (2);

[0070]

[0071] Among them, ε(M) represents the value of the loss function; μ represents the preset coefficient; SKNN(x i ) indicates that x i The set of k nearest neighbors with the same risk category, k is any positive integer; x l Indicates that it belongs to KNN(x i ) set and does not belong to SKNN(x i ) data information of the lth credit card in the set, KNN(x i ) is the same as x i The set of the nearest k neighbors, k is any positive integer; [Z] + =max(Z,0) means the value in the brackets is compared with 0 and the larger one is taken; y il Represents x i With x j The corresponding values ​​for the same or different risk categories, x i With x jWhen the risk categories are the same, the value is 1, x i With x j The value is 0 when the risk categories are different.

[0072] Specifically, an iterative learning process is used to determine the appropriate mapping matrix M representing the original data space in the form of Markov distance, that is, to obtain the Markov matrix M. Assume that the sample set x = {x1, x2, ..., x n}, where x n Represents the data information of the nth credit card in the sample set, that is, the nth sample. The goal of iterative learning is to learn a Markov distance metric through the LMNN algorithm based on n samples with known labels, that is, formula (1) shown above.

[0073] The Markov matrix M can be a semi-positive symmetric matrix and can be expressed as: Where Q represents a matrix, P represents another matrix, and the Markov matrix M can be represented by the product of the transposed matrix of the Q matrix and the P matrix. The Q matrix and the P matrix can be obtained by learning.

[0074] For example, when constructing the loss function, for a risk category L i Sample x i , the sample x i The k nearest neighbors of a target are defined as the k nearest neighbors in the sample set that are related to the sample x. i With the same risk category L i , and the k samples closest to it, then we get the sample x i The set of k nearest neighbors SKNN(x i ), where k can be any positive integer, for example, k can be 10. The set SKNN(x i ) can be understood as sample x i The target sample. If x j is x i The target sample can be expressed as x j ∈SKNN(x i ).

[0075] For example, we can construct a y ij ∈{0,1}, is used to represent the relationship between whether the risk categories of all samples in the sample set are the same. For example, the rows and columns in the binary matrix are n samples arranged in the same order, and the element in the binary matrix is ​​y ij ,y ij =1 indicates sample x i and sample x j The risk category is the same, that is, the label is the same, y ij =0 means sample x i and sample xj The risk categories are different, that is, the labels are different.

[0076] When iterative learning is performed through the large-margin nearest neighbor algorithm, the distance metric suitable for the sample set can be learned through the loss function shown in formulas (2) to (4) above. Among them, formulas (3) and (4) can be substituted into formula (2) and can be understood as two components of formula (2). The part ε shown in formula (3) pull (M), which has the effect of penalizing samples with large distances from their target samples; the part ε shown in formula (4) push (M) Its effect is to penalize small distances between input samples that do not have the same label.

[0077] Exemplarily, a convergence condition can be set to determine whether iterative learning has ended. During iterative learning, by adjusting the parameters in the Markov matrix M, such as adjusting the Q matrix and the P matrix, so that the value ε(M) of the loss function calculated after adjustment meets the convergence condition, it can be determined that the iterative learning has ended. At this time, the learned Markov matrix M can be obtained through the adjusted Q matrix and the P matrix.

[0078] It should be noted that μ represents a preset coefficient, which can be understood as a coefficient for weighting the two parts of formula (3) and formula (4). The value of μ can be preset by empirical value or calculated value. [Z] + = max(Z,0), where Z represents the part in brackets in formula (4), i.e. Z = 1 + D M (x i ,x j )-D M (x i ,x l ), after calculating the value of Z, compare the value of Z with 0. If the value of Z is greater than 0, the value of this part is the value of Z. If the value of Z is less than or equal to 0, the value of this part is 0. l Indicates that it belongs to KNN(x i ) set and does not belong to SKNN(x i ) The data information of the lth credit card in the collection can also be expressed as x l ∈KNN(x i )\SKNN(x i ), where KNN(x i ) indicates that x i The set of samples of the nearest k neighbors, the samples in this set are the same as sample x i The k nearest samples, these neighboring samples can be samples x i Samples with the same label can also be samples with different labels. ilThe value of y can be obtained by constructing ij ∈{0,1} binary matrix.

[0079] Based on this, the LMNN algorithm can be used to obtain a learning-based distance measurement matrix through several rounds of iterative learning. Through the LMNN algorithm, high-dimensional data information can be clustered, and then a distance measurement matrix suitable for the distribution characteristics of credit card data information can be obtained. The distance measurement matrix obtained by this method helps to improve the accuracy of fuzzy calculation of the sample set based on the fuzzy rough set model.

[0080] After obtaining a distance metric matrix based on metric learning, the distance between any two samples in the sample set can be calculated based on the distance metric matrix. This distance can also be understood as the similarity between the two samples. For example, if the calculated distance is smaller, the similarity between the two is higher, and if the calculated distance is larger, the similarity between the two is lower. Based on the distance metric matrix obtained through learning, a fuzzy rough set model based on metric learning can be constructed.

[0081] For example, the information decision system based on metric learning can be denoted as IDS= <U,A,L,Δ M >, where U represents the sample set, A represents the set of all data categories included in the sample set, L represents each risk category, which can also be understood as a label set, Δ M represents the distance metric based on the learned distance metric matrix M. For example, the fuzzy power set of the sample set U can be represented by F(U), Indicates that based on Δ M A fuzzy equivalence relation defined on U, where B represents a subset of A. For any sample x i , we can use δ M (x i ) represents the sample x i Under the distance metric matrix M, about the label L i 's neighborhood.

[0082] For any subset B of A, that is, for The fuzzy lower approximation operator can be defined as: The fuzzy upper approximation operator can be defined as: in, The fuzzy lower approximation operator can also be understood as the fuzzy lower approximation or lower approximation, and the fuzzy upper approximation operator can also be understood as the fuzzy upper approximation or upper approximation.

[0083] Here, a represents an element in set A, which can be understood as a data category in A. i(a) represents sample x i Data information under data category a, x j (a) represents sample x j Data information under data category a, It can be understood as sample x i and sample x j The data information of each data category a is calculated based on the distance metric matrix M, which can characterize the sample x i and sample x j Based on the dissimilarity of Can get Among them, ∩ represents the operation of taking the intersection or taking the minimum.

[0084] In the above formula, δ M (x i ) represents the sample x i Based on the neighborhood of M, if x j ∈δ M (x i ), then D M (x i ,x j )≤δ, that is Among them, δ can be understood as the preset domain radius. The fuzzy lower approximation operator of the sample can represent the degree of certainty that the sample belongs to the risk category to which it belongs, and the fuzzy upper approximation operator of the sample can represent the degree of possibility that the sample belongs to the risk category to which it belongs. Among them, the degree of certainty can be understood as the degree of affirmation, and the degree of possibility can be understood as the degree of possibility. Therefore, in order to determine the correlation between the data category and the risk category, the fuzzy lower approximation operator can be used to perform relevant calculations when selecting the target data category for each data category.

[0085] Furthermore, it can be determined that the fuzzy positive region. Represents x i ∈U is the lower approximation of B for each risk category under the distance metric matrix M. Based on the distance metric matrix M, the fuzzy positive region of all samples of B Defined as:

[0086] Based on the distance metric matrix M obtained after metric learning, after defining the fuzzy lower approximation operator and fuzzy positive region, a fuzzy rough set model based on metric learning is constructed.

[0087] Figure 3 Schematic diagram of the process of determining the credit card risk category provided in the embodiment of the present application Figure 2 , this embodiment is in Figure 2 Step S202 is further refined based on the embodiment, such as Figure 3 As shown, step S202 may specifically include the following steps:

[0088] S301, selecting one or more data categories from each data category of a sample set to obtain a first category subset.

[0089] Exemplarily, all data categories included in the sample set may constitute a set of data categories, which may be understood as the entire set of data categories, and any data category may be an element of the entire set, for example, the entire set is represented by set A. The set A may be determined as a first category subset, and any subset of the set A may also be determined as a first category subset.

[0090] S302: Add one or more data categories that are not selected to the first category subset, or remove one or more data categories from the first category subset to obtain a second category subset.

[0091] Exemplarily, after determining the first category subset, one or more elements belonging to set A but not belonging to the first category subset can be added to the first category subset to obtain a second category subset; alternatively, one or more elements in the first category subset can be removed to obtain a second category subset.

[0092] S303, through the fuzzy rough set model, calculate the fuzzy dependency of the data information of each credit card in the sample set under the first category subset on the risk category to obtain the first fuzzy dependency, and calculate the fuzzy dependency of the data information of each credit card in the sample set under the second category subset on the risk category to obtain the second fuzzy dependency.

[0093] For the elements included in the first category subset and the second category subset, the data information corresponding to each sample in the sample set is calculated through the pre-constructed fuzzy rough set model, and the first fuzzy dependency corresponding to the first category subset and the second fuzzy dependency corresponding to the second category subset can be obtained. The fuzzy dependency can be understood as the indicative information that can characterize the correlation between the data category and the risk category.

[0094] S304, determining whether the data category added or removed when obtaining the second category subset is associated with the risk category based on the first fuzzy dependency and the second fuzzy dependency, and determining the data category associated with the risk category as the target data category.

[0095] Exemplarily, the first category subset and the second category subset may be two sets having an intersection, and both sets belong to a set consisting of all data categories included in the sample set.

[0096] For example, Wherein, A represents the set of all data categories included in the sample set, a represents a data category in set A, i.e., an element, and B is a subset of A; Ba represents the set obtained by removing a from set B, which is also a subset of A. If the first category subset can be B, then the second category subset can be Ba; or, if the first category subset can be Ba, then the second category subset can be B. If a is a data category, then after calculating the fuzzy dependency, it can be obtained whether the data category a is associated with the risk category.

[0097] Exemplarily, the association of two or more data categories can also be determined simultaneously. For example, data category a and data category b are removed from the first category subset B, and the second category subset Bab is obtained. After calculating the fuzzy dependency, it can be determined whether the combination of a and b is associated with the risk category.

[0098] In a possible implementation, the following formulas (5) and (6) are satisfied when calculating the fuzzy dependency through the fuzzy rough set model:

[0099]

[0100] Among them, U represents the sample set; x i represents the data information of the i-th credit card in the sample set; L represents each risk category; M represents the distance measurement matrix; B represents the category subset corresponding to the calculated fuzzy dependency; Represents x i ∈U is the fuzzy lower approximation operator based on B for each risk category; Indicates fuzzy positive area; |*| indicates modulus; Indicates fuzzy dependency.

[0101] Substituting formula (6) into formula (5), we can get the fuzzy dependency function for calculating fuzzy dependency: |U| can be understood as the number of samples in the sample set obtained by taking the modulus of the sample set. The multi-label fuzzy dependency function reflects the approximation ability of the fuzzy equivalence class induced by the set including all data categories under the new measurement method to the label set.

[0102] Through the fuzzy rough set model, the fuzzy dependency of the data information of each credit card in the sample set under the first category subset on the risk category is calculated to obtain the first fuzzy dependency, and the fuzzy dependency of the data information of each credit card in the sample set under the second category subset on the risk category is calculated to obtain the second fuzzy dependency. This process can be understood as calculating the corresponding fuzzy dependencies for the first category subset and the second category subset. Afterwards, the correlation of the data categories in the two category subsets that are not in the intersection of the two can be determined based on the fuzzy dependency. If the data category is regarded as a feature, the process can also be understood as judging redundant features and non-redundant features. Redundant features can be understood as data categories that are not associated with risk categories, and non-redundant features can be understood as data categories that are associated with risk categories.

[0103] When determining whether the data categories added or removed when obtaining the second category subset are associated with the risk category based on the first fuzzy dependency and the second fuzzy dependency, the difference between the first fuzzy dependency and the second fuzzy dependency can be compared with a preset threshold.

[0104] For example, the preset threshold value may be any preset value, such as 0. If the difference between the first fuzzy dependency and the second fuzzy dependency is greater than 0, it can be determined that the data category added or removed when obtaining the second category subset is associated with the risk category; if the difference between the first fuzzy dependency and the second fuzzy dependency is less than or equal to 0, it can be determined that the data category added or removed when obtaining the second category subset is not associated with the risk category.

[0105] It should be understood that the preset threshold can also be a preset threshold range, for example (0, 1). If the difference between the first fuzzy dependency and the second fuzzy dependency belongs to (0, 1), it can be determined that the data category added or removed when obtaining the second category subset is associated with the risk category; if the difference between the first fuzzy dependency and the second fuzzy dependency does not belong to (0, 1), it can be determined that the data category added or removed when obtaining the second category subset is not associated with the risk category.

[0106] Exemplarily, when determining whether the data categories added or removed when obtaining the second category subset are associated with the risk category based on the first fuzzy dependency and the second fuzzy dependency, it can be determined by whether the difference between the first fuzzy dependency and the second fuzzy dependency is equal to a preset threshold. The preset threshold is, for example, 0. Then, if the difference between the first fuzzy dependency and the second fuzzy dependency is not equal to 0, it can be determined that the data categories added or removed when obtaining the second category subset are associated with the risk category; if the difference between the first fuzzy dependency and the second fuzzy dependency is equal to 0, it can be determined that the data categories added or removed when obtaining the second category subset are not associated with the risk category.

[0107] Specifically, let IDS = <U,A,L,Δ M > is a decision system based on metric learning, Δ M is the distance metric matrix M, For any data category a, let a∈A and a∈B, if Right now When , a is indispensable for B. It can be understood that after removing data category a from the first category subset B, the second category subset Ba is obtained. At this time, if the calculated first fuzzy dependency of the first category subset is not equal to the calculated second fuzzy dependency of the second category subset, it indicates that the data information of each sample in the sample set under data category a has an influence on the calculation of the lower approximation of the label set, then a is indispensable for B. Therefore, the data category a can have an impact on the determination of the risk category, which further indicates that the data category a and the risk category are related. The data category a can be understood as a non-redundant feature. Similarly, the difference between the first fuzzy dependency and the second fuzzy dependency is equal to 0, that is When , it indicates that the data information of each sample in the sample set under data category a has no effect when calculating the lower approximation of the label set, then a is not essential for B, and the data category a can be understood as a redundant feature.

[0108] The essential non-redundant features can provide necessary information support for the risk category classification of credit cards. However, redundant features not only fail to provide effective information for risk category classification, but may also reduce the performance of the classifier and prolong the classification time. Therefore, through the embodiment of the present application, the correlation between data categories and risk categories is determined, and the screening of each data category is achieved. When the target data category is determined, the data category that is useful for risk category classification can be retained, which helps to better divide the sample space during classification, and then build a better performance machine learning model for classification prediction.

[0109] Exemplary, metric learning-based decision system IDS= <U,A,L,Δ M >, the feature dependency of data categories can be defined, which can characterize the strength of the correlation between the corresponding data categories and risk categories. If B is a category subset of A, for any a∈AB, that is, for any data category in A but not in B, its feature dependency can be expressed as: Among them, Sig(a,B) represents the feature dependency of data category a. represents the fuzzy dependency calculated under the category subset B∪{a}, Represents the fuzzy dependency calculated under category subset B.

[0110] From the expression of feature dependency, we can see that if 0≤Sig(a,B)≤1, when Sig(a,B)=0, it means that a is a redundant feature for B, and when Sig(a,B)≠0, it means that a is a non-redundant feature for B. The larger the Sig(a,B), the higher the importance of data category a, that is, the stronger the correlation between data category a and the risk category, and the stronger the distinguishing ability, then when data category a is used to distinguish the risk categories of credit cards, the higher the accuracy of the determined risk category results.

[0111] For example, for a given IDS= <U,A,L,Δ M >, if Then B can be understood as a subset of C if condition (1) is satisfied: And satisfy condition (2): For any a∈B, , it means that when the number of data categories in set B is less than the number of data categories in set C, and the distinguishing abilities of set C and set B are the same, set B can be called a simplification of set C. At this time, set B can be understood as a category subset that includes several target data categories with higher distinguishing abilities, and the number of target data categories in the set is minimized. Based on this simplified category subset, when determining the risk category of the credit card to be tested, data information corresponding to data categories with higher correlation can be selected from the data information of credit cards with more dimensions, thereby effectively reducing the amount of input data used for calculation.

[0112] In the embodiment of the present application, since the fuzzy rough set model is a mathematical model combining fuzzy set theory and rough set theory, it can be used to deal with uncertainty and imprecision. On the basis of determining this specific application scenario in combination with the credit card risk category, by constructing a fuzzy rough set model based on metric learning, it can better adapt to the distribution characteristics of data information in the sample set of credit cards, realize effective distance measurement and correlation judgment, and thus improve the accuracy of determining the target data category.

[0113] S203, when at least one target data category is determined, inputting data information of the credit card to be tested under at least one target data category into a classification model for classification prediction to determine the risk category of the credit card to be tested.

[0114] Exemplarily, the classification model can be any machine model that can perform classification tasks after pre-training, such as a linear model, a support vector machine (SVM) model and / or a decision tree, etc. For the credit card to be tested whose risk category is to be determined, the data information under the target data category after the data category is simplified is input into the classification model, and classification prediction can be performed to determine the risk category of the credit card to be tested.

[0115] In one possible implementation, data information of the credit card to be tested under at least one target data category can be input into a classification model to calculate the function value of the credit card to be tested corresponding to the target function; classification prediction is performed based on the function value and a preset judgment threshold to determine the risk category of the credit card to be tested; wherein the target function is used to calculate the probability of the risk category corresponding to the credit card to be tested, the target function is preset based on the support vector distance between the risk category of the credit card in the sample set and the high-dimensional discriminant function, the high-dimensional discriminant function is obtained by transforming the linear discriminant function through the Gaussian kernel, and the linear discriminant function is determined based on the data information in the sample set using a support vector machine model under at least one target data category.

[0116] Exemplarily, the SVM model may be trained using the reduced category subset to determine the risk category of the credit card to be tested using the trained SVM model.

[0117] For the sample set x={x1,x2,…,x n}, any sample can be expressed as (x i ,L i ), i=1,2,3,……n,L i represents the risk category of the i-th sample, L∈{+1,-1}, for example, L i =1 indicates risk. x∈ in, The data information representing the sample is data information belonging to d dimensions, where d dimensions correspond to the number of at least one determined target data category. For example, if 10 target data categories are determined, then d is 10.

[0118] Based on the data information of each sample in the sample set and the labels corresponding to the samples, the SVM model is trained to determine the hyperplane (w, b) used for classification. In the d-dimensional space, the linear discriminant function can be expressed as: g(x) = w·x+b. By deforming x through the kernel function, the linear discriminant function in the d-dimensional space can be transformed into a high-dimensional discriminant function: g(x) = w·φ(x)+b.

[0119] Since the dimensionality of the samples in the sample set is high, if the selected kernel function is not suitable for the high-dimensional characteristics of the sample set, there may be a problem of insufficient dimensionality. Therefore, the kernel function parameters can be limited to the Gaussian kernel in the radial basis function kernel (RBF). Since the Gaussian kernel can map the original feature space to infinite dimensions, it can better adapt to the high-dimensional characteristics of the data information in the sample set. The expression of the Gaussian kernel can be: in, Represents the high-dimensional feature distribution parameter, which determines the dispersion of samples after the original data is mapped to the high-dimensional space. This will cause the samples to be more sparsely dispersed. The farther apart the samples are, the easier they are to separate. However, while extreme samples are more easily separated, the model is also more likely to overfit. x and y represent the representation of the same sample before and after the transformation, respectively.

[0120] For example, the high-dimensional discriminant function is normalized so that the samples of the two risk categories satisfy the absolute value |g(x)|≥1, the closest sample |g(x)|=1, and the distance between the support vector and the discriminant surface is 2 / ∥w∥, so that the maximum classification interval is equivalent to:

[0121] Exemplarily, considering the case of linear inseparability, a relaxation vector ξ and a penalty coefficient c can be introduced on the basis of the above embodiment, where c represents the tolerance to errors. If the sample point falls on the correct side of the maximum margin hyperplane, the relaxation variable is 0<ξ i <1; for misclassified sample point j, the relaxation vector ξ j＝1 The relaxation vector is multiplied by the penalty coefficient c. When c is large, even a small relaxation vector will have a significant impact on the final result. The objective function can then be transformed into: After solving the objective function and the data information of the credit card to be tested, the probability representing the risk category corresponding to the credit card to be tested can be obtained. The probability can be the function value of the objective function.

[0122] Classification prediction is performed based on the function value and the preset judgment threshold to determine the risk category of the credit card to be tested. The judgment threshold may be a preset value for judging the function value, such as 0.5 or any other value.

[0123] For example, if the judgment condition is defined as the function value is greater than or equal to the judgment threshold, the risk category is risky, and if the function value is less than the judgment threshold, the risk category is risk-free. For example, if the function value of the objective function is 0.7 for the data information of a credit card to be tested, and the function value is greater than the judgment threshold of 0.5, it can be determined that the risk category corresponding to the credit card to be tested is risky. Risky can be understood as the risk of credit card fraud and other behaviors.

[0124] Based on this, the data information of the credit card to be tested under at least one target data category is input into the classification model to calculate the function value of the credit card to be tested corresponding to the target function; classification prediction is performed according to the function value and the preset judgment threshold to determine the risk category of the credit card to be tested. Based on this, the risk category of the credit card to be tested can be determined more accurately and efficiently.

[0125] The credit card risk category determination method provided by the embodiment of the present application obtains a sample set through the data information of the credit card marked with the risk category, and determines which data categories in the sample set are associated with the risk category through the fuzzy rough set model, and determines the target data category associated with the risk category. Since the fuzzy rough set model can determine the association according to the fuzzy approximate operator, it can accurately screen out the data category that is helpful to determine the risk category from the numerous data categories, so as to achieve the purpose of filtering out the data category that is irrelevant to the determination of the risk category, and can reduce the number of data categories, and further reduce the amount of information of the data information. When at least one target data category is determined, the data information of the credit card to be tested under at least one target data category is input into the classification model for classification prediction. Compared with the case where the data information under all the data categories of the credit card to be tested is input, this method can reduce the amount of data input into the classification model, reduce the amount of calculation when the classification model is calculated, and further shorten the time required for calculation, thereby improving the efficiency of classification prediction. Since the filtered data category is an irrelevant data category, the interference of irrelevant data information is eliminated, so that the accuracy of the prediction result is higher. Therefore, based on this method, the effect of improving the determination efficiency can be achieved while taking into account the accuracy of determining the credit card risk category.

[0126] In addition, in the financial field, the interpretability of the model is more important, and the decision-making basis of the model needs to be explained to regulators and users. In the above-mentioned related technologies, since deep learning models are generally considered to be black box models, their decision-making process lacks transparency, and the interpretability of the models in the prior art is poor. By constructing a fuzzy rough set model to determine the target data category, the mathematical expression of the fuzzy rough set model is more interpretable, so the method based on the embodiment of the present application can improve the interpretability of the risk category determination process.

[0127] Next, combine Figure 4 A specific implementation of the credit card risk category determination method provided in an embodiment of the present application is described. Figure 4 Schematic diagram of the process of determining the credit card risk category provided in the embodiment of the present application Figure 3 ,like Figure 4 As shown, the method includes:

[0128] S401, collecting data information of multiple credit cards, pre-processing each data information, and converting it into a feature vector.

[0129] Exemplarily, when preprocessing data information, it may be a data cleaning operation such as completing and correcting errors in the data information, or it may include normalizing the data information to eliminate differences in data information dimensions between different data categories.

[0130] For discrete data in data information, binary numbers such as 0 or 1 can be assigned to make it digital. Discrete data can be information that represents yes or no. For example, when using the distance metric Δ M When the data information a under the i-th data category i When x(a i )≠y(a i ), that is, when the data information of sample x and sample y in the i-th data category is not equal, x(a i )-y(a i )=1; when x(a i )=y(a i ), that is, when the data information of sample x and sample y in the i-th data category is equal, x(a i )-y(a i )=0.

[0131] For example, the data information of all data categories of a credit card can be connected and represented in the form of a vector, which can be converted into the feature vector of the credit card. The feature vector can be understood as the multi-dimensional vector mentioned above, such as (XXXX XXXX XXXX, 200.01, ...).

[0132] S402, constructing a distance measurement function suitable for the collected sample set through the LMNN algorithm.

[0133] For example, the distance metric function may be constructed by using the methods in the above embodiments. The distance metric function may be, for example, formula (1) in the above embodiments.

[0134] S403, constructing an ML-FRS model.

[0135] Exemplarily, by constructing the ML-FRS model, a fuzzy upper approximation operator, a fuzzy lower approximation operator and / or a fuzzy positive region etc. can be defined, and then a function for calculating the fuzzy dependency can be obtained, such as formula (5) and formula (6) in the above embodiment.

[0136] S404, determining the target data category based on the ML-FRS model and the sample set.

[0137] Exemplarily, a traversal calculation can be performed for each data category included in the sample set to determine whether each data category is a data category associated with the risk category, and then the target data category can be comprehensively determined. For example, for any data category a∈B, by constructing a first category subset B and a second category subset Ba, a fuzzy dependency calculation is performed through the ML-FRS model to determine whether the data category a is associated with the risk category, so as to determine whether it is the target data category.

[0138] S405, classifying the credit cards to be tested by using the SVM model according to the target data category.

[0139] After the target data category is determined, for each target data category, the data information of the credit card to be tested under each target data category can be processed by the SVM model to predict the risk category of the credit card to be tested.

[0140] The method provided in the embodiment of the present application can perform feature selection on the data categories in the sample set, screen out the features that are more suitable for identifying whether it is a credit card use risk, that is, determine the target data category. Based on this, when there are many data categories for credit cards, the dimension of data information can be reduced, and the calculation efficiency of model prediction can be improved. In practical applications, this method can be used to identify the risks of credit card fraud such as false applications and stolen cards. By verifying the applicant's information, analyzing the transaction pattern, and other means, abnormal situations can be discovered in a timely manner to protect the financial security of banks and cardholders.

[0141] Figure 5 A schematic diagram of the structure of a credit card risk category determination device provided in an embodiment of the present application, such as Figure 5 As shown, an embodiment of the present application provides a credit card risk category determination device, the device comprising:

[0142] The acquisition module 501 is used to acquire a sample set, which includes data information of n credit cards, each of which is marked with its own risk category, and the data information of the n credit cards corresponds to m data categories. Any credit card includes data information under m data categories, and m and n are both integers greater than 1.

[0143] A determination module 502 is used to determine whether each data category in the sample set is associated with the risk category through a fuzzy rough set model, and determine the data category associated with the risk category as the target data category. The fuzzy rough set model is a model that determines the association based on a fuzzy approximation operator;

[0144] The processing module 503 is used to input the data information of the credit card to be tested under the at least one target data category into the classification model for classification prediction to determine the risk category of the credit card to be tested when at least one target data category is determined.

[0145] In one possible implementation, the determination module 502 is specifically used to: select one or more data categories from each data category of the sample set to obtain a first category subset; add one or more unselected data categories to the first category subset, or remove one or more data categories from the first category subset to obtain a second category subset; calculate the fuzzy dependency of the data information of each credit card in the sample set under the first category subset on the risk category through a fuzzy rough set model to obtain a first fuzzy dependency, and calculate the fuzzy dependency of the data information of each credit card in the sample set under the second category subset on the risk category to obtain a second fuzzy dependency; determine whether the data categories added or removed when obtaining the second category subset are associated with the risk category based on the first fuzzy dependency and the second fuzzy dependency, and determine the data categories associated with the risk category as target data categories.

[0146] In a possible implementation, the following formula is satisfied when the fuzzy dependency is calculated by the fuzzy rough set model: Among them, U represents the sample set; x i represents the data information of the i-th credit card in the sample set; L represents each risk category; M represents the distance measurement matrix; B represents the category subset corresponding to the calculated fuzzy dependency; Represents x i ∈U is the fuzzy lower approximation operator based on B for each risk category; Indicates fuzzy positive area; |*| indicates modulus; Indicates fuzzy dependency.

[0147] In a possible implementation, the distance metric matrix is ​​determined after iterative learning through a large-margin nearest neighbor algorithm based on data information in a sample set.

[0148] In a possible implementation, the distance metric matrix is ​​a Markov matrix, and the Markov matrix satisfies the following formula: Where M represents the Markov matrix; x j represents the data information of the jth credit card in the sample set; D M (x i ,x j ) represents x i With x j Markov distance; when iterative learning is performed through the large margin nearest neighbor algorithm, the loss function used satisfies the following formula: ε(M) = (1-μ)ε pull (M)+με push (M); Among them, ε(M) represents the value of the loss function; μ represents the preset coefficient; SKNN(x i ) indicates that x iThe set of k nearest neighbors with the same risk category, k is any positive integer; x l Indicates that it belongs to KNN(x i ) set and does not belong to SKNN(x i ) data information of the lth credit card in the set, KNN(x i ) is the same as x i The set of the nearest k neighbors, k is any positive integer; [Z] + =max(Z,0) means the value in the brackets is compared with 0 and the larger one is taken; y il Represents x i With x j The corresponding values ​​for the same or different risk categories, x i With x j When the risk categories are the same, the value is 1, x i With x j The value is 0 when the risk categories are different.

[0149] In one possible implementation, the processing module 503 is specifically used to: input data information of the credit card to be tested under at least one target data category into a classification model to calculate the function value of the credit card to be tested corresponding to the objective function; perform classification prediction based on the function value and a preset judgment threshold to determine the risk category of the credit card to be tested; wherein the objective function is used to calculate the probability of the risk category corresponding to the credit card to be tested, the objective function is preset based on the support vector distance between the risk category of the credit card in the sample set and the high-dimensional discriminant function, the high-dimensional discriminant function is obtained by transforming the linear discriminant function through the Gaussian kernel, and the linear discriminant function is determined based on the data information in the sample set using a support vector machine model under at least one target data category.

[0150] The credit card risk category determination device provided in the embodiment of the present application can be used to execute the technical solution of the credit card risk category determination method in any of the above embodiments of the present application. Its implementation principle and technical effect are similar, and this embodiment will not be repeated here.

[0151] Figure 6 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application is shown in FIG. Figure 6 As shown, the electronic device of this embodiment may include: at least one processor 601; and a memory 602 communicatively connected to the at least one processor; wherein the memory 602 stores instructions executable by the at least one processor 601, and the instructions are executed by the at least one processor 601 so that the electronic device executes a method as in any of the above embodiments.

[0152] Optionally, the memory 602 may be independent or integrated with the processor 601 .

[0153] The implementation principle and technical effects of the electronic device provided in this embodiment can be found in the aforementioned embodiments and will not be described in detail here.

[0154] An embodiment of the present application further provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer-executable instructions. When a processor executes the computer-executable instructions, the method described in any of the above embodiments is implemented.

[0155] An embodiment of the present application further provides a computer program product, including a computer program, which implements the method described in any of the above embodiments when executed by a processor.

[0156] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of the modules is only a logical function division, and there may be other division methods in actual implementation, such as multiple modules can be combined or integrated into another system, or some features can be ignored or not executed.

[0157] The above-mentioned integrated module implemented in the form of a software function module can be stored in a computer-readable storage medium. The above-mentioned software function module is stored in a storage medium, including a number of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor to perform some steps of the method described in each embodiment of the present application.

[0158] It should be understood that the above-mentioned processor can be a processing unit (Central Processing Unit, CPU), or other general-purpose processors, digital signal processors (Digital Signal Processor, DSP), application-specific integrated circuits (Application Specific Integrated Circuit, ASIC), etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc. The steps of the method disclosed in the application can be directly embodied as being executed by a hardware processor, or can be executed by a combination of hardware and software modules in the processor. The memory may include random access memory (Random Access Memory, RAM), and may also include non-volatile memory (Non-Volatile Memory, NVM), such as at least one disk storage, and can also be a USB flash drive, a mobile hard disk, a read-only memory, a disk or an optical disk, etc.

[0159] The above storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read only memory (EEPROM), erasable programmable read only memory (EPROM), programmable read only memory (PROM), read only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk. The storage medium can be any available medium that can be accessed by a general or special-purpose computer.

[0160] An exemplary storage medium is coupled to a processor so that the processor can read information from the storage medium and write information to the storage medium. Of course, the storage medium can also be a component of the processor. The processor and the storage medium can be located in a dedicated integrated circuit. Of course, the processor and the storage medium can also exist as discrete components in an electronic device or a main control device.

[0161] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0162] The serial numbers of the above-mentioned embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

[0163] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in each embodiment of the present application.

[0164] The above are only preferred embodiments of the present application, and are not intended to limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the present application specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present application.

[0165] Those skilled in the art will readily come up with other implementations of the embodiments of the present application after considering the specification and practicing the application disclosed herein. The present application is intended to cover any modifications, uses or adaptations of the embodiments of the present application, which follow the general principles of the embodiments of the present application and include common knowledge or customary technical means in the art that are not disclosed in the embodiments of the present application.

[0166] It should be understood that the embodiments of the present application are not limited to the precise structures described above and shown in the drawings, and various modifications and changes can be made without departing from the scope thereof. The scope of the embodiments of the present application is limited only by the appended claims.

Claims

1. A method for determining a credit card risk category, characterized in that: include: Obtain a sample set, the sample set including data information of n credit cards, the n credit cards are all marked with respective risk categories, the data information of the n credit cards correspond to m data categories, any of the credit cards includes data information under the m data categories, and both m and n are integers greater than 1; Determine whether each data category in the sample set is associated with the risk category by using a fuzzy rough set model, and determine the data category associated with the risk category as the target data category, wherein the fuzzy rough set model is a model for determining the association based on a fuzzy approximation operator; When at least one target data category is determined, data information of the credit card to be tested under the at least one target data category is input into a classification model for classification prediction to determine the risk category of the credit card to be tested.

2. The method according to claim 1, characterized in that The method of determining whether each data category in the sample set is associated with a risk category by using a fuzzy rough set model, and determining the data category associated with the risk category as the target data category, includes: Select one or more data categories from each data category of the sample set to obtain a first category subset; Adding one or more data categories that are not selected to the first category subset, or removing one or more data categories from the first category subset, to obtain a second category subset; By using the fuzzy rough set model, the fuzzy dependency of the data information of each credit card in the sample set under the first category subset on the risk category is calculated to obtain a first fuzzy dependency, and the fuzzy dependency of the data information of each credit card in the sample set under the second category subset on the risk category is calculated to obtain a second fuzzy dependency; According to the first fuzzy dependency and the second fuzzy dependency, it is determined whether the data category added or removed when obtaining the second category subset is associated with the risk category, and the data category associated with the risk category is determined as the target data category.

3. The method according to claim 2, characterized in that The following formula is satisfied when calculating fuzzy dependency through fuzzy rough set model: Wherein, U represents the sample set; x i represents the data information of the i-th credit card in the sample set; L represents each risk category; M represents the distance measurement matrix; B represents the category subset corresponding to the calculated fuzzy dependency; Represents x i ∈U is the fuzzy lower approximation operator based on B for each risk category; Indicates fuzzy positive area; |*| indicates modulus; Indicates fuzzy dependency.

4. The method according to claim 3, characterized in that: The distance metric matrix is ​​determined after iterative learning through a large-margin nearest neighbor algorithm based on the data information in the sample set.

5. The method according to claim 4, characterized in that The distance metric matrix is ​​a Markov matrix, and the Markov matrix satisfies the following formula: Wherein, M represents the Markov matrix; x j represents the data information of the jth credit card in the sample set; D M (x i ,x j ) represents x i With x j The Markov distance of When the large-margin nearest neighbor algorithm is used for iterative learning, the loss function used satisfies the following formula: ε(M)=(1-μ)ε pull (M)+with push (M); Among them, ε(M) represents the value of the loss function; μ represents the preset coefficient; SKNN(x i ) indicates that x i The set of k nearest neighbors with the same risk category, k is any positive integer; x l Indicates that it belongs to KNN(x i ) set and does not belong to SKNN(x i ) data information of the lth credit card in the set, KNN(x i ) is the same as x i The set of the nearest k neighbors, k is any positive integer; [Z] + =max(Z,0) means the value in the brackets is compared with 0 and the larger one is taken; y il Represents x i With x j The corresponding values ​​for the same or different risk categories, x i With x j When the risk categories are the same, the value is 1, x i With x j The value is 0 when the risk categories are different.

6. The method according to any one of claims 1 to 5, characterized in that: The step of inputting data information of the credit card to be tested under the at least one target data category into a classification model for classification prediction to determine the risk category of the credit card to be tested includes: Inputting data information of the credit card to be tested under the at least one target data category into a classification model to calculate a function value of the credit card to be tested corresponding to a target function; Performing classification prediction based on the function value and a preset determination threshold to determine the risk category of the credit card to be tested; Among them, the objective function is used to calculate the probability of the risk category corresponding to the credit card to be tested, and the objective function is preset based on the risk category of the credit cards in the sample set and the high-dimensional discriminant function through the support vector distance. The high-dimensional discriminant function is obtained by transforming the linear discriminant function through the Gaussian kernel, and the linear discriminant function is determined under the at least one target data category using the support vector machine model based on the data information in the sample set.

7. A credit card risk category determination device, characterized in that: include: an acquisition module, used to acquire a sample set, wherein the sample set includes data information of n credit cards, each of the n credit cards is marked with a respective risk category, the data information of the n credit cards corresponds to m data categories, any of the credit cards includes data information under the m data categories, and both m and n are integers greater than 1; A determination module, used to determine whether each data category in the sample set is associated with the risk category through a fuzzy rough set model, and determine the data category associated with the risk category as the target data category, wherein the fuzzy rough set model is a model for determining the association based on a fuzzy approximation operator; The processing module is used to input the data information of the credit card to be tested under the at least one target data category into the classification model for classification prediction to determine the risk category of the credit card to be tested.

8. An electronic device, characterized in that: include: Memory and processor; The memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory, so that the processor performs the method according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, which are used to implement the method according to any one of claims 1 to 6 when executed by a processor.

10. A computer program product, characterized in that The method comprises a computer program, which implements the method according to any one of claims 1 to 6 when being executed by a processor.

Images