Credit card fraud detection method and device based on graph neural network, and storage medium

Abstract

The invention provides a credit card fraud detection method and device based on a graph neural network and a storage medium. The credit card fraud detection method based on the graph neural network comprises the steps of obtaining to-be-processed data; inputting to-be-processed data into the trained credit card fraud detection model to obtain the category of the to-be-processed data; the credit card fraud detection model is used for classifying input data into normal data or fraud data, and the credit card fraud detection model is formed based on a graph neural network and a hybrid expert network. According to the credit card fraud detection method and device based on the graph neural network and the storage medium provided by the invention, the credit card fraud detection model obtained based on training of the graph neural network and the hybrid expert network can more fully learn information in the input data, so that a more accurate detection result is obtained.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a credit card fraud detection method, device and storage medium based on graph neural network. Background Art

[0002] Credit card payment is a common payment method in our daily life, widely used in many occasions such as eating, shopping, traveling, etc. However, with the popularity of credit cards, credit cards may be used by criminals to commit fraud and defraud property. Fraudulent transactions may cause serious losses to the cardholder company and the credit card company. Therefore, using transaction information to detect fraud helps card issuers to achieve anti-fraud and protect the property safety of cardholders.

[0003] However, due to the privacy of users and transaction secrets, it is difficult to obtain real credit card transaction information, or only very little transaction information can be obtained, and the quality of the obtained data is not good enough, resulting in poor credit card fraud detection results and inability to ensure transaction security. Summary of the invention

[0004] The present invention provides a credit card fraud detection method, device and storage medium based on graph neural network, which are used to solve the technical problem of inaccurate credit card fraud detection in the prior art.

[0005] In a first aspect, the present invention provides a credit card fraud detection method based on a graph neural network, comprising the following steps: Get the data to be processed; The data to be processed is input into a trained credit card fraud detection model to obtain the category of the data to be processed; the credit card fraud detection model is used to classify the input data into normal data or fraud data, and the credit card fraud detection model is based on a graph neural network and a hybrid expert network.

[0006] In some embodiments, the credit card fraud detection model is trained by the following steps: Associating the data to be tested with the sample data; Determine the association relationship between the test data and the sample data based on a hybrid expert network; The credit card fraud detection model is iteratively trained based on the association relationship and using a graph convolutional network.

[0007] In some embodiments, the determining the association relationship between the test data and the sample data based on the hybrid expert network includes: Determine the relationship category between the test data and the sample data based on the hybrid expert network, and characterize the association relationship between the test data and the sample data in the form of weight; The relationship categories between the data to be tested and the sample data include information supplementation relationship and information interference relationship.

[0008] In some embodiments, associating the data to be tested with the sample data includes: Clustering the sample data to obtain multiple training sample clusters; Acquire a training sample graph based on the sample data and the cluster center of the training sample cluster; The data to be tested is connected to the cluster center of each training sample cluster in the training sample graph.

[0009] In some embodiments, the step of acquiring a training sample graph based on the sample data and the cluster center of the training sample cluster includes: The sample data and the cluster centers of the training sample clusters are taken as nodes, and undirected edges are added between the nodes belonging to the same training sample cluster, so as to obtain a structural diagram corresponding to each training sample cluster; The structural graphs corresponding to multiple training sample clusters are connected by connecting the nodes of different clusters to obtain a training sample graph.

[0010] In some embodiments, the endpoints in the training sample graph satisfy: the difference between the edge density of the endpoints belonging to different training sample clusters and the edge density of the endpoints belonging to the same training sample cluster is less than a preset threshold.

[0011] In some embodiments, the iterative training of the credit card fraud detection model based on the association relationship and using a graph convolutional network includes: Based on the association relationship and using the forward propagation of the graph convolutional network, the node vector of the data to be tested in the training sample graph is updated; Determine the predicted category of the data to be tested according to the updated node vector, and determine the cross entropy loss function according to the actual category of the data to be tested and the predicted category; Model parameters of the credit card fraud detection model are optimized based on the cross entropy loss function and the updated node vector using back propagation.

[0012] In a second aspect, the present invention provides a credit card fraud detection device based on a graph neural network, comprising the following modules: An acquisition module is used to acquire data to be processed; A classification module is used to input the data to be processed into a trained credit card fraud detection model to obtain the category of the data to be processed; the credit card fraud detection model is used to classify the input data into normal data or fraud data, and the credit card fraud detection model is based on a graph neural network and a hybrid expert network.

[0013] In a third aspect, the present invention provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method for detecting credit card fraud based on a graph neural network as described in the first aspect above is implemented.

[0014] In a fourth aspect, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the credit card fraud detection method based on graph neural network as described in the first aspect above.

[0015] In a fifth aspect, the present invention further provides a computer program product, comprising a computer program, which, when executed by a processor, implements the credit card fraud detection method based on graph neural network as described in the first aspect above.

[0016] The credit card fraud detection method, device and storage medium based on graph neural network provided by the present invention obtain the data to be processed, input the data to be processed into a trained credit card fraud detection model, and obtain the category of the data to be processed, wherein the credit card fraud detection model is used to classify the input data into normal data or fraud data, and the credit card fraud detection model is based on graph neural network and hybrid expert network, so that the credit card fraud detection model can learn the information in the input data more fully, thereby obtaining more accurate detection results. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0018] Figure 1 It is a flow chart of the credit card fraud detection method based on graph neural network provided by the present invention.

[0019] Figure 2 It is a structural schematic diagram of the credit card fraud detection model provided by the present invention.

[0020] Figure 3 It is a structural schematic diagram of a credit card fraud detection device based on graph neural network provided by the present invention.

[0021] Figure 4 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0022] Due to the privacy issues of credit card usage, current datasets used to train credit card fraud detection models cannot provide original features and more contextual information about the data. This data problem exists not only in credit card scenarios, but also in other areas of life. In the data-driven era, data is becoming more and more valuable, and data sharing and protection have become a serious issue.

[0023] Models based on deep learning rely on a large amount of training data, and a serious lack of data will lead to overfitting of the model. Due to the high cost of acquisition and labeling, samples with labeled information are actually very rare. How to use a small amount of data set to train an accurate credit card fraud detection model and achieve better detection results even with small sample learning is a problem that needs to be solved.

[0024] Based on the above technical problems, the present invention proposes a credit card fraud detection method based on graph neural network, which uses a credit card fraud detection model composed of graph neural network and hybrid expert network to process the data to be processed, associates the data through graph neural network, and optimizes the data flow through hybrid expert network to help fully and effectively utilize the information carried by the data, thereby improving the efficiency and accuracy of model training and improving detection performance.

[0025] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0026] Figure 1 is a flow chart of a credit card fraud detection method based on graph neural network provided by the present invention, such as Figure 1 As shown, the present invention provides a credit card fraud detection method based on graph neural network. The method includes the following steps 101 and 102.

[0027] Step 101: Obtain data to be processed.

[0028] Specifically, the data to be processed refers to data to be classified, which may include user transaction information and transaction behavior data, etc.

[0029] Step 102: input the data to be processed into a trained credit card fraud detection model to obtain the category of the data to be processed; the credit card fraud detection model is used to classify the input data into normal data or fraud data, and the credit card fraud detection model is based on a graph neural network and a hybrid expert network.

[0030] Specifically, the credit card fraud detection model is based on Graph Neural Networks (GNN) and Mixture of Experts (MoE), which is used to identify whether the input data is normal data or fraudulent data. When applied, the data to be processed is input into the trained credit card fraud detection model for classification, and the category of the data to be processed is output.

[0031] The credit card fraud detection method based on graph neural network provided in the embodiment of the present application, and the credit card fraud detection model obtained by training based on graph neural network and hybrid expert network, can more fully learn the information in the input data, thereby obtaining more accurate detection results.

[0032] In some embodiments, the credit card fraud detection model is trained by the following steps: Associate the data to be tested with the sample data one by one; Determine the association relationship between the test data and the sample data based on a hybrid expert network; The credit card fraud detection model is iteratively trained based on the association relationship and using a graph convolutional network.

[0033] Specifically, the data to be tested is test data with unknown labels, and the model needs to determine its category; but due to the semi-supervised training paradigm of graph neural networks, this part of the data is visible during training and is also used to assist in the training of credit card fraud detection models; the sample data is labeled training data, and the labels of the training samples include two categories, one is normal data and the other is fraud data.

[0034] To train the credit card fraud detection model, the test data is first associated with the sample data. For example, a graph structure is constructed based on the test data and the sample data. Then, the association relationship between the test data and the sample data is determined based on the hybrid expert network, and the graph convolution network is used to optimize the network parameters through iterative training, and finally a trained credit card fraud detection model is obtained.

[0035] In some embodiments, associating the data to be tested with the sample data includes: Clustering the sample data to obtain multiple training sample clusters; Acquire a training sample graph based on the sample data and the cluster center of the training sample cluster; The data to be tested is connected to the cluster center of each training sample cluster in the training sample graph.

[0036] Specifically, the data to be tested is associated with the sample data, and the sample data may be clustered first to obtain a plurality of training sample clusters.

[0037] For example, the K-means clustering method is used to cluster all sample data according to the characteristics of the sample data to generate multiple training sample clusters.

[0038] Then, a training sample graph is obtained based on the sample data and the cluster centers of the training sample clusters. For example, sample data in the same training sample cluster are connected in pairs, cluster centers of the training sample clusters are connected in pairs, and sample data of some different clusters are selected for connection to obtain a training sample graph.

[0039] In some embodiments, the step of acquiring a training sample graph based on the sample data and the cluster center of the training sample cluster includes: The sample data and the cluster centers of the training sample clusters are taken as nodes, and undirected edges are added between the nodes belonging to the same training sample cluster, so as to obtain a structural diagram corresponding to each training sample cluster; The structural graphs corresponding to multiple training sample clusters are connected by connecting the nodes of different clusters to obtain a training sample graph.

[0040] Specifically, to obtain a training sample graph, the sample data and the cluster centers of all training sample clusters can be constructed into nodes, and unweighted undirected edges can be constructed between nodes belonging to the same training sample cluster to obtain a structural graph corresponding to each training sample cluster. Then, the structural graphs corresponding to multiple training sample clusters are connected by connecting nodes of different clusters, specifically including connecting the cluster centers of the training sample clusters in pairs, selecting nodes belonging to different clusters and connecting them (the selection method can be a method of randomly selecting nodes with uniform distribution), and finally obtaining a training sample graph.

[0041] In some embodiments, the endpoints in the training sample graph satisfy: the difference between the edge density of the endpoints belonging to different training sample clusters and the edge density of the endpoints belonging to the same training sample cluster is less than a preset threshold.

[0042] Specifically, in the training sample graph, it is necessary to ensure that the edge density of endpoints belonging to different training sample clusters is roughly equal to the edge density of endpoints belonging to the same training sample cluster. This can be done by setting a smaller threshold. When the difference between the edge density of endpoints belonging to different training sample clusters and the edge density of endpoints belonging to the same training sample cluster is less than the preset threshold, it is determined that the two are roughly equal and the condition is met. If the difference between the two is greater than or equal to the preset threshold, it is necessary to further adjust the training sample graph or regenerate the training sample graph. The validity of the training sample graph is guaranteed, thereby ensuring the validity and accuracy of subsequent model training based on the training sample graph.

[0043] After generating a training sample graph based on the sample data, the data to be tested is connected to the cluster center of each training sample cluster in the training sample graph, thereby realizing the association between the data to be tested and the sample data.

[0044] The credit card fraud detection method based on graph neural network provided in the embodiment of the present application associates the data to be tested with the sample data by clustering and constructing a graph structure, so that the training sample data can be effectively used to guide the detection of the data to be tested.

[0045] In some embodiments, the determining the association relationship between the test data and the sample data based on the hybrid expert network includes: Determine the relationship category between the test data and the sample data based on the hybrid expert network, and characterize the association relationship between the test data and the sample data in the form of weight; The relationship categories between the data to be tested and the sample data include information supplementation relationship and information interference relationship.

[0046] Specifically, the relationship category between the test data and the sample data is determined based on the hybrid expert network, that is, the connection edges between the test data and the sample data are classified, and it is determined whether the relationship between the data to be processed and the training samples belongs to an information supplement relationship or an information interference relationship, and the degree or strength of this association relationship can be represented in the form of the weight of the connection edge.

[0047] The relationship categories between the data to be tested and the sample data include information supplementation relationship and information interference relationship. If there is a strong information supplementation relationship between the data to be processed and the training samples, training the credit card fraud detection model based on the data to be processed and the training samples can improve the training efficiency and accuracy of the model, which is conducive to improving the detection performance; if there is a strong information interference relationship between the data to be processed and the training samples, training the credit card fraud detection model based on the data to be processed and the training samples together will affect the training efficiency and accuracy of the model.

[0048] For example, a hybrid expert network is constructed, which consists of a routing network and four expert networks. The routing network is composed of a multilayer perceptron (MLP). The routing network determines which expert(s) should be assigned to each pair of nodes (i.e., the node pair consisting of the test data and the sample data) for processing. The first two expert networks are responsible for modeling the information complement relationship between node pairs, and the last two expert networks are responsible for modeling the information interference relationship between node pairs. The expert network outputs a weight value, which represents the degree of association between two nodes.

[0049] The credit card fraud detection method based on graph neural network provided in the embodiment of the present application utilizes a hybrid expert network to obtain the correlation between the data to be tested and the sample data, optimizes the data flow through a routing network, and determines the correlation between each pair of nodes more flexibly and efficiently, thereby achieving better detection performance and accuracy.

[0050] In some embodiments, the iterative training of the credit card fraud detection model based on the association relationship and using a graph convolutional network includes: Based on the association relationship and using the forward propagation of the graph convolutional network, the node vector of the data to be tested in the training sample graph is updated; Determine the predicted category of the data to be tested according to the updated node vector, and determine the cross entropy loss function according to the actual category of the data to be tested and the predicted category; Model parameters of the credit card fraud detection model are optimized based on the cross entropy loss function and the updated node vector using back propagation.

[0051] Specifically, the association relationship between the data to be tested and the sample data is obtained, that is, the weight of the connection between the nodes corresponding to the data to be tested and the nodes corresponding to the sample data is obtained, and the node vector corresponding to the data to be tested is updated using the forward propagation of the graph convolutional network to obtain an updated training sample graph; the predicted category of the data to be tested is determined according to the updated node vector, and the cross entropy loss function is calculated in combination with the actual category of the data to be tested.

[0052] If the number of iterations does not reach the preset number of times, or the value of the obtained cross entropy loss function does not meet the preset conditions, back propagation is used to optimize the model parameters of the credit card fraud detection model, a new round of iterations is started using the updated training sample graph, and the cross entropy loss function corresponding to the new round of iteration results is calculated until the preset number of iterations is completed, or the calculated cross entropy loss function meets the conditions, the training is stopped to obtain a trained credit card fraud detection model.

[0053] The information transmission methods provided in the above embodiments are further described below through specific examples: Example 1: Figure 2 Schematic diagram of the structure of the credit card fraud detection model provided by the present invention. Figure 2 As shown in the figure, the specific training steps for the credit card fraud detection model are as follows: Step 1: Normalize all sample data, and use the K-means clustering method to cluster all available training sample data according to sample characteristics to form K training sample clusters, and calculate the corresponding clusters.

[0054] Step 2: Construct a training sample graph: construct all sample data and K cluster centers into nodes, construct unweighted undirected edges between nodes belonging to the same cluster, connect the K cluster centers in pairs, and then connect nodes belonging to different clusters by randomly selecting nodes in a uniform distribution, while ensuring that the edge density of endpoints belonging to different clusters is roughly equal to the edge density of endpoints belonging to the same cluster.

[0055] Step 3: Iterate and execute the following steps 4 to 7.

[0056] Step 4: Add a sample to be tested to the training sample graph and connect it to the K cluster center nodes to form a semi-supervised learning training-detection meta-task.

[0057] Step 5: Use the hybrid expert network to determine the information supplement effect between the two nodes on each edge, and use the output of the hybrid expert network to supplement the weight of the edge:

[0058] in, represents the weight of the edge between the i-th node and the j-th node in the k-th layer network output by the hybrid expert network (i and j are not equal), which represents the information supplement / interference relationship between the two nodes. represents the i-th node representation of the input k-th layer network, Represents the j-th node representation of the input k-th layer network. It represents a hybrid expert network, which consists of a routing network and four experts. The routing network R is composed of a multi-layer perceptron (MLP), which is responsible for determining which expert should be assigned to each pair of nodes for judgment:

[0059] In the formula, r Representative experts, there are four experts 1, 2, 3, and 4; represents the concatenation operation, The nodes representing bitwise operations represent the absolute value of the difference. Each expert is composed of a shallower MLP, which is responsible for mapping the node pair data into the network edge weights. Among them, the first two experts are responsible for modeling the information complement relationship between node pairs, and the last two experts are responsible for modeling the information interference relationship between node pairs:

[0060] In the formula, Represents the weight of the edge between the i-th node and the j-th node in the k-th layer network output by the hybrid expert network (i and j are not equal). r When the weight is 1 or 2, it is positive, indicating that there is an information complement relationship between the two nodes.r When it is 3 or 4, the weight value is negative, indicating that there is an information interference relationship between the two nodes; Represents the activation function sigmoid().

[0061] Step 6: Use the graph convolutional network (Graph Conv in the figure) to propagate forward to update the node vector ( Updated to ), determine the predicted category of the data to be tested based on the updated node vector, and use an activation function (i.e., the Softmax Function in the figure) to output the classification result as the corresponding probability value of the category 0 or 1 (i.e., the value of probability p, the category is 1, indicating that the predicted category of the data to be tested is normal data, and the value is 0, indicating that the predicted category of the data to be tested is fraudulent data). Calculate the cross entropy loss function (i.e., the Loss Function in the figure), then update the network parameters through back propagation, and perform training for this meta-task by iteratively performing this process for multiple rounds; Step 7: If all meta-task training is completed, the iteration ends, otherwise return to step 4.

[0062] The credit card fraud detection method based on graph neural network provided in the embodiment of the present application associates the test samples one by one with all the training samples through a graph, constructs a learnable parameter adjacency matrix method, relies on the similarity of features to pass messages and uses the graph convolutional network layer to extract node features, so that the model can effectively detect credit card fraud when there is very little training data. At the same time, this method uses a hybrid expert system to fully model the information supplement relationship between samples that is conducive to improving detection performance and the information interference relationship that is not conducive to improving detection performance. Therefore, compared with other similar methods, this application can perform better when there is an extreme shortage of training samples.

[0063] Figure 3 is a schematic diagram of the structure of a credit card fraud detection device based on a graph neural network provided by the present invention, such as Figure 3 As shown, the present invention provides a credit card fraud detection device based on graph neural network, including an acquisition module 301 and a classification module 302.

[0064] The acquisition module 301 is used to acquire data to be processed.

[0065] The classification module 302 is used to input the data to be processed into the trained credit card fraud detection model to obtain the category of the data to be processed; the credit card fraud detection model is used to classify the input data into normal data or fraud data, and the credit card fraud detection model is based on a graph neural network and a hybrid expert network.

[0066] In some embodiments, the credit card fraud detection model is trained by the following steps: Associating the data to be tested with the sample data; Determine the association relationship between the test data and the sample data based on a hybrid expert network; The credit card fraud detection model is iteratively trained based on the association relationship and using a graph convolutional network.

[0067] In some embodiments, the determining the association relationship between the test data and the sample data based on the hybrid expert network includes: Determine the relationship category between the test data and the sample data based on the hybrid expert network, and characterize the association relationship between the test data and the sample data in the form of weight; The relationship categories between the data to be tested and the sample data include information supplementation relationship and information interference relationship.

[0068] In some embodiments, associating the data to be tested with the sample data includes: Clustering the sample data to obtain multiple training sample clusters; Acquire a training sample graph based on the sample data and the cluster center of the training sample cluster; The data to be tested is connected to the cluster center of each training sample cluster in the training sample graph.

[0069] In some embodiments, the step of acquiring a training sample graph based on the sample data and the cluster center of the training sample cluster includes: The sample data and the cluster centers of the training sample clusters are taken as nodes, and undirected edges are added between the nodes belonging to the same training sample cluster, so as to obtain a structural diagram corresponding to each training sample cluster; The structural graphs corresponding to multiple training sample clusters are connected by connecting the nodes of different clusters to obtain a training sample graph.

[0070] In some embodiments, the endpoints in the training sample graph satisfy: the difference between the edge density of the endpoints belonging to different training sample clusters and the edge density of the endpoints belonging to the same training sample cluster is less than a preset threshold.

[0071] In some embodiments, the iterative training of the credit card fraud detection model based on the association relationship and using a graph convolutional network includes: Based on the association relationship and using the forward propagation of the graph convolutional network, the node vector of the data to be tested in the training sample graph is updated; Determine the predicted category of the data to be tested according to the updated node vector, and determine the cross entropy loss function according to the actual category of the data to be tested and the predicted category; Model parameters of the credit card fraud detection model are optimized based on the cross entropy loss function and the updated node vector using back propagation.

[0072] Specifically, the above-mentioned credit card fraud detection device based on graph neural network provided by the present invention can implement all the method steps implemented by the above-mentioned credit card fraud detection method embodiment based on graph neural network, and can achieve the same technical effect. The parts and beneficial effects of this embodiment that are the same as the method embodiment will not be described in detail here.

[0073] It should be noted that the division of units / modules in the above-mentioned embodiments of the present invention is schematic and is only a logical function division. There may be other division methods in actual implementation. In addition, the functional units in the various embodiments of the present application may be integrated into a processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated units may be implemented in the form of hardware or in the form of software functional units.

[0074] Figure 4 is a schematic diagram of the structure of the electronic device provided by the present invention, such as Figure 4 As shown, the electronic device may include: a processor 401, a communications interface 402, a memory 403 and a communication bus 404, wherein the processor 401, the communications interface 402 and the memory 403 communicate with each other through the communication bus 404. The processor 401 may call the logic instructions in the memory 403 to execute a credit card fraud detection method based on a graph neural network, the method comprising: Get the data to be processed; The data to be processed is input into a trained credit card fraud detection model to obtain the category of the data to be processed; the credit card fraud detection model is used to classify the input data into normal data or fraud data, and the credit card fraud detection model is based on a graph neural network and a hybrid expert network.

[0075] Specifically, the processor 401 may be a central processing unit (CPU), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or a complex programmable logic device (CPLD), and the processor may also adopt a multi-core architecture.

[0076] The logic instructions in the memory 403 can be implemented in the form of software functional units and can be stored in a processor-readable storage medium when sold or used as an independent product. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) or a processor to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program codes.

[0077] In some embodiments, a computer program product is further provided, the computer program product comprising a computer program, the computer program may be stored on a non-transitory computer-readable storage medium, and when the computer program is executed by a processor, the computer can execute the credit card fraud detection method based on graph neural network provided by the above-mentioned method embodiments, the method comprising: Get the data to be processed; The data to be processed is input into a trained credit card fraud detection model to obtain the category of the data to be processed; the credit card fraud detection model is used to classify the input data into normal data or fraud data, and the credit card fraud detection model is based on a graph neural network and a hybrid expert network.

[0078] Specifically, the above-mentioned computer program product provided in the embodiment of the present application can implement all the method steps implemented by the above-mentioned method embodiments, and can achieve the same technical effect. The parts and beneficial effects of this embodiment that are the same as the method embodiment will not be described in detail here.

[0079] In some embodiments, a computer-readable storage medium is further provided, wherein the computer-readable storage medium stores a computer program, wherein the computer program is used to enable a computer to execute the credit card fraud detection method based on graph neural network provided in the above-mentioned method embodiments, and the method comprises: Get the data to be processed; The data to be processed is input into a trained credit card fraud detection model to obtain the category of the data to be processed; the credit card fraud detection model is used to classify the input data into normal data or fraud data, and the credit card fraud detection model is based on a graph neural network and a hybrid expert network.

[0080] Specifically, the above-mentioned computer-readable storage medium provided by the present invention can implement all the method steps implemented by the above-mentioned method embodiments, and can achieve the same technical effects. The parts and beneficial effects that are the same as the method embodiments in this embodiment will not be described in detail here.

[0081] It should be noted that the computer-readable storage medium can be any available medium or data storage device that can be accessed by the processor, including but not limited to magnetic storage (such as floppy disks, hard disks, magnetic tapes, magneto-optical disks (MO), etc.), optical storage (such as CD, DVD, BD, HVD, etc.), and semiconductor storage (such as ROM, EPROM, EEPROM, non-volatile memory (NAND FLASH), solid-state drive (SSD)), etc.

[0082] In the present invention, the term "plurality" refers to two or more than two, and other quantifiers are similar to them.

[0083] In the present invention, "determine B based on A" means that the factor A should be considered when determining B. It is not limited to "B can be determined based on A alone", but should also include: "determine B based on A and C", "determine B based on A, C and E", "determine C based on A, and further determine B based on C", etc. In addition, it can also include taking A as a condition for determining B, for example, "when A meets the first condition, use the first method to determine B"; for another example, "when A meets the second condition, determine B", etc.; for another example, "when A meets the third condition, determine B based on the first parameter", etc. Of course, it can also be a condition that takes A as a factor for determining B, for example, "when A meets the first condition, use the first method to determine C, and further determine B based on C", etc.

[0084] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage and optical storage, etc.) containing computer-usable program codes.

[0085] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer executable instructions. These computer executable instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0086] These processor executable instructions may also be stored in a processor readable memory that can direct a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the processor readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0087] These processor-executable instructions may also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for executing on the computer or other programmable device to implement the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0088] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A credit card fraud detection method based on graph neural network, characterized in that: include: Get the data to be processed; Inputting the data to be processed into a trained credit card fraud detection model to obtain the category of the data to be processed; The credit card fraud detection model is used to classify input data into normal data or fraud data, and the credit card fraud detection model is based on a graph neural network and a hybrid expert network.

2. The credit card fraud detection method based on graph neural network according to claim 1, characterized in that: The credit card fraud detection model is trained by the following steps: Associating the data to be tested with the sample data; Determine the association relationship between the test data and the sample data based on a hybrid expert network; The credit card fraud detection model is iteratively trained based on the association relationship and using a graph convolutional network.

3. The credit card fraud detection method based on graph neural network according to claim 2 is characterized in that: The determining the association relationship between the test data and the sample data based on the hybrid expert network includes: Determine the relationship category between the test data and the sample data based on the hybrid expert network, and characterize the association relationship between the test data and the sample data in the form of weight; The relationship categories between the data to be tested and the sample data include information supplementation relationship and information interference relationship.

4. The credit card fraud detection method based on graph neural network according to claim 2, characterized in that: The associating the data to be tested with the sample data includes: Clustering the sample data to obtain multiple training sample clusters; Acquire a training sample graph based on the sample data and the cluster center of the training sample cluster; The data to be tested is connected to the cluster center of each training sample cluster in the training sample graph.

5. The credit card fraud detection method based on graph neural network according to claim 4 is characterized in that: The step of acquiring a training sample graph based on the sample data and the cluster center of the training sample clustering clusters includes: The sample data and the cluster centers of the training sample clusters are taken as nodes, and undirected edges are added between the nodes belonging to the same training sample cluster, so as to obtain a structural diagram corresponding to each training sample cluster; The structural graphs corresponding to multiple training sample clusters are connected by connecting the nodes of different clusters to obtain a training sample graph.

6. The credit card fraud detection method based on graph neural network according to claim 4 or 5, characterized in that: The endpoints in the training sample graph satisfy that: the difference between the edge density of the endpoints belonging to different training sample clusters and the edge density of the endpoints belonging to the same training sample cluster is less than a preset threshold.

7. The credit card fraud detection method based on graph neural network according to claim 5, characterized in that: The iterative training of the credit card fraud detection model based on the association relationship and using a graph convolutional network includes: Based on the association relationship and using the forward propagation of the graph convolutional network, the node vector of the data to be tested in the training sample graph is updated; Determine the predicted category of the data to be tested according to the updated node vector, and determine the cross entropy loss function according to the actual category of the data to be tested and the predicted category; Model parameters of the credit card fraud detection model are optimized based on the cross entropy loss function and the updated node vector using back propagation.

8. A credit card fraud detection device based on graph neural network, characterized in that: include: An acquisition module is used to acquire data to be processed; A classification module is used to input the data to be processed into a trained credit card fraud detection model to obtain the category of the data to be processed; the credit card fraud detection model is used to classify the input data into normal data or fraud data, and the credit card fraud detection model is based on a graph neural network and a hybrid expert network.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the credit card fraud detection method based on graph neural network as described in any one of claims 1 to 7 is implemented.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for detecting credit card fraud based on a graph neural network as described in any one of claims 1 to 7 is implemented.

Images