Fraudulent transaction account detection method based on heterogeneous graph convolution network

Abstract

The application provides a fraud transaction account detection method based on a heterogeneous graph convolution network, which comprises the following steps: constructing a bipartite graph by taking a plurality of target transaction accounts as account nodes, taking a plurality of orders associated with each target transaction account as order nodes, and taking the association relationship between each target transaction account and each order associated with the target transaction account as an order edge; wherein the order edge comprises an order submission edge and an order cancellation edge; performing convolution processing on the bipartite graph based on the heterogeneous graph convolution network to obtain an account node hidden state corresponding to the account node, an order submission edge hidden state and an order cancellation edge hidden state; and inputting the account node hidden state, the order submission edge hidden state and the order cancellation edge hidden state into a pre-acquired classification model to make the classification model output account anomaly detection result data of the account node. The application can effectively improve the accuracy and detection efficiency of transaction fraud account detection.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of fraudulent transaction account detection, and particularly relates to a fraudulent transaction account detection method based on a heterogeneous graph convolution network. BACKGROUND

[0002] The carbon trading market mechanism is a new system for controlling and reducing greenhouse gas emissions through market mechanisms. It is widely recognized as an effective, reliable and transparent policy tool for reducing carbon emissions, and is an important tool for achieving the dual carbon goal. Both domestic and international practices have proven that carbon trading not only enables enterprises to recognize the importance of carbon emission reduction, but also drives other subjects to participate in voluntary emission reduction through economic incentives, and is a method for reducing the overall emission reduction cost at a lower cost. The carbon trading market is the fastest growing commodity market in the world. According to the World Bank, the estimated value of the carbon trading market will reach 176 billion US dollars in the future, and this huge market is under the threat of unstable policies and lack of regulation. The EU carbon market is the largest and longest-running carbon emissions trading market in the world. Since the start of the EU carbon market, a series of carbon trading manipulation and fraud cases have occurred, not only causing huge economic losses to victims and related member states, but also posing a serious challenge to the EU carbon market regulatory system. A fair market trading environment is an important prerequisite for promoting efficient resource allocation and achieving price discovery function. However, fraudulent trading behavior in the carbon market not only undermines the fairness of carbon market trading, but also causes abnormal fluctuations in carbon quota prices and deviates from the true value determined by market supply and demand, causing huge losses to investors while greatly increasing the financial risks of the carbon market.

[0003] The existing various transaction scenarios include the detection method of account node fraudulent transaction behavior in the carbon trading scenario, which has the problems of poor accuracy and low efficiency. SUMMARY

[0004] In view of this, the embodiments of the present application provide a fraudulent transaction account detection method based on a heterogeneous graph convolution network to eliminate or improve one or more defects in the prior art.

[0005] A first aspect of the present application provides a fraudulent transaction account detection method based on a heterogeneous graph convolution network, the method comprising:

[0006] A plurality of target transaction accounts are respectively taken as account nodes, a plurality of orders respectively associated with each of the target transaction accounts are respectively taken as order nodes, and an association relationship between each of the target transaction accounts and each of the orders respectively associated therewith is taken as an order edge, to construct a bipartite graph; wherein the order edge comprises an order submission edge and an order cancellation edge;

[0007] obtaining an account node hidden state, an order submission edge hidden state and an order cancellation edge hidden state corresponding to the account node by performing convolutional processing on the bipartite graph based on the heterogeneous graph convolutional network;

[0008] inputting the account node hidden state, the order submission edge hidden state and the order cancellation edge hidden state into a pre-obtained classification model, so that the classification model outputs account abnormality detection result data of the account node.

[0009] In some embodiments of the present application, the target transaction account is a carbon transaction account, and the order is a carbon transaction order.

[0010] The account node hidden state, the order submission edge hidden state and the order cancellation edge hidden state corresponding to the account node are obtained by performing convolutional processing on the bipartite graph based on the heterogeneous graph convolutional network, including:

[0011] obtaining an order node hidden state corresponding to each order node by performing convolutional processing on each order node based on the heterogeneous graph convolutional network;

[0012] obtaining an order edge hidden state corresponding to each order edge by performing convolutional processing on each order edge based on the heterogeneous graph convolutional network;

[0013] obtaining the order submission edge hidden state based on each order node hidden state, each order edge hidden state, a combination function, a first aggregation function and an order submission edge training weight matrix;

[0014] obtaining the order cancellation edge hidden state based on each order node hidden state, each order edge hidden state, the combination function, a second aggregation function and an order cancellation edge training weight matrix;

[0015] obtaining the account node hidden state based on the order submission edge hidden state and the order cancellation edge hidden state of the previous layer, the account node hidden state, the combination function, an account node training weight matrix and a third aggregation function.

[0016] In some embodiments of the present application, the order node hidden state corresponding to each order node is obtained by performing convolutional processing on each order node based on the heterogeneous graph convolutional network, including:

[0017] obtaining the order node hidden state corresponding to each order node based on the account node hidden state and the order edge hidden state of the previous layer corresponding to each order node, an order node training weight matrix and the combination function.

[0018] In some embodiments of the present application, the order edge hidden state corresponding to each of the order edges is obtained based on the convolution processing of each of the order edges by the heterogeneous graph convolution network, including:

[0019] The order edge hidden state corresponding to each of the order edges is obtained based on the account node hidden state, the order node hidden state, the order edge hidden state and the order edge training weight matrix of the previous layer corresponding to each of the order edges.

[0020] In some embodiments of the present application, the order submission edge hidden state is obtained based on the order node hidden state, the order edge hidden state, the combination function, the first aggregation function and the order submission edge training weight matrix, and the corresponding formula is as follows:

[0021]

[0022]

[0023] wherein, represents the order submission edge hidden state, and σ represents a nonlinear activation function, represents the order submission edge training weight matrix, represents the first aggregation function, represents the order node hidden state, represents the order edge hidden state, a represents an account node, o represents an order node, and concat represents the combination function.

[0024] In some embodiments of the present application, the order cancellation edge hidden state is obtained based on the order node hidden state, the order edge hidden state, the combination function, the second aggregation function and the order cancellation edge training weight matrix, and the corresponding formula is as follows:

[0025]

[0026]

[0027] wherein, represents the order cancellation edge hidden state, and σ represents a nonlinear activation function, represents the order cancellation edge training weight matrix, represents the second aggregation function, represents the order node hidden state, represents the order edge hidden state, a represents an account node, o represents an order node, and concat represents the combination function.

[0028] In some embodiments of the present application, the account node hidden state is obtained based on the order submission edge hidden state of the previous layer, the order cancellation edge hidden state of the previous layer, the account node hidden state of the previous layer, the combination function, the account node training weight matrix and the third aggregation function, and the corresponding formula is as follows:

[0029]

[0030]

[0031] wherein, represents the account node hidden state, and σ represents a nonlinear activation function, represents the account node training weight matrix, represents the third aggregation function, represents the account node hidden state of the previous layer, represents the order submission edge hidden state of the previous layer, represents the order cancellation edge hidden state of the previous layer, and concat represents the combination function.

[0032] The second aspect of the present application provides a computer device, comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor executes the computer program to realize the fraud transaction account detection method based on the heterogeneous graph convolution network according to the first aspect.

[0033] The third aspect of the present application provides a computer readable storage medium, which stores a computer program, wherein the computer program is executed by a processor to realize the fraud transaction account detection method based on the heterogeneous graph convolution network according to the first aspect.

[0034] The fourth aspect of the present application provides a computer program product, comprising a computer program, wherein the computer program is executed by a processor to realize the fraud transaction account detection method based on the heterogeneous graph convolution network according to the first aspect.

[0035] The application provides a fraud transaction account detection method based on a heterogeneous graph convolution network, which comprises the following steps: constructing a bipartite graph by taking a plurality of target transaction accounts as account nodes, taking a plurality of orders associated with each target transaction account as order nodes, and taking the association relationship between each target transaction account and each order associated with the target transaction account as an order edge; wherein the order edge comprises an order submission edge and an order cancellation edge; performing convolution processing on the bipartite graph based on a heterogeneous graph convolution network to obtain an account node hidden state, an order submission edge hidden state and an order cancellation edge hidden state corresponding to the account node; and inputting the account node hidden state, the order submission edge hidden state and the order cancellation edge hidden state into a pre-acquired classification model to make the classification model output an account abnormality detection result data of the account node. The application can effectively improve the accuracy and detection efficiency of transaction fraud account detection.

[0036] Additional advantages, objects, and features of the application will be set forth in part in the description which follows, and in part will become apparent to those having ordinary skill in the art upon examination of the following or can be learned from practice of the application. The objects and other advantages of the application can be realized and attained by the structure particularly pointed out in the written description and claims hereof as well as the appended drawings.

[0037] It will be understood by those skilled in the art that the objects and advantages of the present application can be realized and attained by the structure particularly pointed out in the written description and claims hereof as well as the appended drawings. BRIEF DESCRIPTION OF DRAWINGS

[0038] The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this application, illustrate embodiments of the application and together with the description serve to explain the principles of the application. The components in the drawings are not drawn to scale, but are merely intended to illustrate the principles of the application. In order to facilitate the illustration and description of some parts of the application, the corresponding parts in the drawings can be enlarged, i.e., can become larger than other components in the exemplary device actually manufactured according to the application. In the drawings:

[0039] Figure 1 The flowchart of the fraud transaction account detection method in an embodiment of the application.

[0040] Figure 2 The structure diagram of the bipartite graph in another embodiment of the application.

[0041] Figure 3 The overall architecture diagram of the fraud transaction account detection method in an embodiment of the application. DETAILED DESCRIPTION

[0042] For the purposes of the present application, the technical solutions and advantages thereof will be more clearly apparent from the following detailed description of embodiments and from the attached drawings. Hereinafter, the illustrative embodiments of the present application and their description serve the purpose of explanations of the present application, but are not intended to limit the present application.

[0043] It should also be noted that, in order not to obscure the present application with unnecessary details, only structures and / or processing steps closely related to the solutions according to the present application are shown in the drawings, and other details not closely related to the present application are omitted.

[0044] It should be emphasized that the terms "comprises / comprising" when used in this specification, specify the presence of stated features, elements, steps or components, but do not preclude the presence or addition of one or more other features, elements, steps or components.

[0045] It should also be noted that, unless otherwise specified, the term "connected" herein can not only mean direct connection, but also indirect connection in the presence of an intermediate.

[0046] Hereinafter, embodiments of the present application will be described with reference to the accompanying drawings. In the drawings, the same reference numerals represent the same or similar parts, or the same or similar steps.

[0047] The present application is specifically described in detail through the following embodiments.

[0048] The embodiment of the present application provides a fraud transaction account detection method based on a heterogeneous graph convolution network, which can be executed by a client device, referring to Figure 1 , the fraud transaction account detection method based on the heterogeneous graph convolution network specifically comprises the following contents:

[0049] Step 110: Constructing a bipartite graph by taking a plurality of target transaction accounts as account nodes, respectively, taking a plurality of orders associated with each of the target transaction accounts as order nodes, respectively, and taking the association relationship between each of the target transaction accounts and each of the orders associated with each of the target transaction accounts as an order edge. The order edge includes an order submission edge and an order cancellation edge.

[0050] Step 120: performing convolution processing on the bipartite graph based on a heterogeneous graph convolution network to obtain an account node hidden state corresponding to the account node, an order submission edge hidden state, and an order cancellation edge hidden state.

[0051] Step 130: inputting the account node hidden state, the order submission edge hidden state, and the order cancellation edge hidden state into a pre-acquired classification model, so that the classification model outputs an account anomaly detection result data of the account node.

[0052] Specifically, the client device first constructs a bipartite graph (as shown in Figure 2 ) by taking a plurality of target transaction accounts as account nodes, taking a plurality of orders associated with each target transaction account as order nodes, and taking the association between each target transaction account and each order associated therewith as an order edge, wherein the order edge includes an order submission edge and an order cancellation edge.

[0053] Then, the bipartite graph is convoluted based on a heterogeneous graph convolutional network (i.e., heterogeneous GCN in Figure 3 ) to obtain an account node hidden state corresponding to the account node, an order submission edge hidden state, and an order cancellation edge hidden state.

[0054] Finally, the account node hidden state, the order submission edge hidden state, and the order cancellation edge hidden state are input into a pre-obtained classification model (i.e., classifier in Figure 3 ), which is a multi-layer perceptron (mlp), so that the classification model outputs an account anomaly detection result data (e.g., normal user and fraud implementer in Figure 3 ), thereby effectively improving the accuracy and efficiency of transaction fraud account detection.

[0055] It should be noted that the method can be applied to various transaction scenarios, such as carbon transaction scenarios, and the corresponding target transaction accounts are carbon transaction accounts, and the orders are carbon transaction orders.

[0056] In order to effectively extract the multi-element features of the transaction account nodes, step 120 includes:

[0057] Step 121: convoluting each order node based on the heterogeneous graph convolutional network to obtain an order node hidden state corresponding to each order node.

[0058] Step 122: convoluting each order edge based on the heterogeneous graph convolutional network to obtain an order edge hidden state corresponding to each order edge.

[0059] Step 123: obtaining the order submission edge hidden state based on each order node hidden state, each order edge hidden state, a combination function, a first aggregation function, and an order submission edge training weight matrix.

[0060] Step 124: obtaining the order cancellation edge hidden state based on each order node hidden state, each order edge hidden state, the combination function, a second aggregation function, and an order cancellation edge training weight matrix.

[0061] Step 125: obtaining the account node hidden state based on the order submission edge hidden state of the previous layer, the order cancellation edge hidden state, the account node hidden state, the combination function, the account node training weight matrix and the third aggregation function.

[0062] Specifically, the client device performs convolution processing on each order node based on the heterogeneous graph convolution network to obtain an order node hidden state corresponding to each order node; performs convolution processing on each order edge based on the heterogeneous graph convolution network to obtain an order edge hidden state corresponding to each order edge; obtains an order submission edge hidden state based on each order node hidden state, each order edge hidden state, a combination function, a first aggregation function and an order submission edge training weight matrix; obtains an order cancellation edge hidden state based on each order node hidden state, each order edge hidden state, a combination function, a second aggregation function and an order cancellation edge training weight matrix; and obtains an account node hidden state based on the order submission edge hidden state of the previous layer, the order cancellation edge hidden state, the account node hidden state, the combination function, the account node training weight matrix and the third aggregation function, so as to effectively extract the multi-element features of the transaction account node.

[0063] In order to further obtain the order node hidden state, step 121 comprises:

[0064] Based on the account node hidden state of the previous layer corresponding to each order node, the order node hidden state, the order edge hidden state and the order node training weight matrix, an order node hidden state corresponding to each order node is obtained. The corresponding formula is shown in formula 1.

[0065]

[0066] wherein, represents the order node hidden state of the L-1th propagation layer in the heterogeneous graph convolution network, represents the account node hidden state of the L-1th propagation layer in the heterogeneous graph convolution network, represents the order edge hidden state of the L-1th propagation layer in the heterogeneous graph convolution network, and concat represents the combination function, represents the order node training weight matrix.

[0067] In order to further obtain the order node hidden state, step 122 comprises:

[0068] Based on the account node hidden state of the previous layer corresponding to each order edge, the order node hidden state, the order edge hidden state and the order edge training weight matrix, an order edge hidden state corresponding to each order edge is obtained. The corresponding formula is shown in formula 2.

[0069]

[0070] wherein, represents the order node hidden state of the L-1th propagation layer in the heterogeneous graph convolution network, represents the account node hidden state of the L-1th propagation layer in the heterogeneous graph convolution network, represents the order edge hidden state of the L-1th propagation layer in the heterogeneous graph convolution network, and concat represents a combination function, represents the order edge training weight matrix. In order to further obtain the order submission edge hidden state, the corresponding formula is shown in equation 3:

[0071]

[0072]

[0073] wherein, represents the order submission edge hidden state, and σ represents a nonlinear activation function, represents the order submission edge training weight matrix, represents a first aggregation function, represents the order node hidden state, represents the order edge hidden state, a represents an account node, o represents an order node, and concat represents a combination function.

[0074] In addition, a self-attention mechanism is added in the order submission edge hidden state, as shown in equation 4:

[0075]

[0076] wherein, ATTN S is a function f: maps the feature vector h key and the candidate feature vector set to the weighted sum of the elements in . The weight of the sum, i.e. the attention value, is calculated by scaling the dot product attention.

[0077] In order to further obtain the order cancellation edge hidden state, the corresponding formula of step 124 is shown in equation 5:

[0078]

[0079]

[0080] wherein, represents the order cancellation edge hidden state, and σ represents a nonlinear activation function, represents the order cancellation edge training weight matrix, represents a second aggregation function, denotes the order node hidden state, denotes the order edge hidden state, a denotes the account node, o denotes the order node, and concat denotes the combination function.

[0081] In addition, a self-attention mechanism is added to the order cancellation edge hidden state, as shown in equation 6:

[0082]

[0083] where ATTN C is a function f: maps the feature vector h key and the candidate feature vector set to the weighted sum of the elements in The weight of the sum, i.e., the attention value, is calculated by scaling the dot product attention.

[0084] In order to further obtain the account node hidden state, step 125 corresponds to the formula as shown in equation 7:

[0085]

[0086]

[0087] where denotes the account node hidden state, and sigma denotes a nonlinear activation function, denotes the account node training weight matrix, denotes a third aggregation function, denotes the account node hidden state of the previous layer, denotes the order submission edge hidden state of the previous layer, denotes the order cancellation edge hidden state of the previous layer, and concat denotes the combination function.

[0088] In addition, compared with the heterogeneous graph convolutional network of the present application, most of the traditional graph convolutional network-based work focuses on homogeneous graphs, assuming is a homogeneous graph, nodes v∈V, and edges The feature of node v is where d0 denotes the feature dimension of the node. The hidden state of node v learned by the model at the l-th layer is denoted as d l The dimension of the l-th layer hidden state is denoted as d. The GCN-based method follows a hierarchical propagation manner, and all nodes are updated simultaneously in each propagation layer. A propagation layer can be divided into two sub-layers: aggregation and combination. Generally speaking, for an L-layer GCN, the aggregation sub-layer and the combination sub-layer of the l-th layer (l=1, 2,..., L) are shown in equations 8 and 9:

[0089]

[0090]

[0091] where N(v) is the set of nodes adjacent to v, AGG is a function used to aggregate the neighbor embeddings of node v, which can be customized by the specific model, e.g. max-pooling, mean-pooling or attention-based weighted sum. W l is a trainable matrix shared by all nodes in layer l. σ is a nonlinear activation function, e.g. Relu. denotes the aggregated features of node v in the neighborhood of layer l. The COMBINE function is used to combine the self-embedding and the aggregated embeddings of neighbors, which is also a customization for different graph models, e.g. the common concatenation. In GCN and GAT, there is no explicit combination sublayer. By introducing the self-information of v in 8 instead of N(v), Thus, the COMBINE step actually happens inside the AGG step.

[0092] In summary, the application provides a fraud transaction account detection method based on a heterogeneous graph convolutional network, which comprises: taking a plurality of target transaction accounts as account nodes, respectively, taking a plurality of orders associated with each of the target transaction accounts as order nodes, respectively, and taking the association relationship between each of the target transaction accounts and each of the orders associated with each of the target transaction accounts as an order edge, and constructing a bipartite graph; wherein the order edge comprises an order submission edge and an order cancellation edge; performing convolutional processing on the bipartite graph based on a heterogeneous graph convolutional network to obtain an account node hidden state corresponding to the account node, an order submission edge hidden state, and an order cancellation edge hidden state; inputting the account node hidden state, the order submission edge hidden state, and the order cancellation edge hidden state into a pre-acquired classification model to make the classification model output an account abnormality detection result data of the account node. The application can effectively improve the accuracy and detection efficiency of transaction fraud account detection.

[0093] The application also provides an electronic device, such as a central server, which can include a processor, a memory, a receiver, and a transmitter. The processor is used to execute the fraud transaction account detection method mentioned in the above embodiments. The processor and the memory can be connected through a bus or other means to be connected through the bus. The receiver can be connected to the processor and the memory through wired or wireless means.

[0094] ​The processor can be a central processing unit (CPU). The processor can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic components, discrete hardware components, or a combination thereof.

[0095] The memory, as a non-transitory computer readable storage medium, can be used to store non-transitory software programs, non-transitory computer executable programs and modules, such as program instructions / modules corresponding to the fraud transaction account detection method in the embodiments of the present application. The processor executes various functions and data processing of the processor by running the non-transitory software programs, instructions and modules stored in the memory, that is, implements the fraud transaction account detection method in the above method embodiments.

[0096] The memory can include a program storage area and a data storage area. The program storage area can store an operating system and at least one application required by a function. The data storage area can store data created by the processor and the like. In addition, the memory can include a high-speed random access memory, and can also include a non-transitory memory, such as at least one magnetic disk storage device, a flash memory device, or other non-transitory solid-state memory device. In some embodiments, the memory can optionally include a memory remotely arranged with respect to the processor, and these remote memories can be connected to the processor through a network. Examples of the above network include but are not limited to the Internet, an intranet, a local area network, a mobile communication network, and a combination thereof.

[0097] The one or more modules are stored in the memory and, when executed by the processor, perform the fraud transaction account detection method in the embodiments.

[0098] In some embodiments of the present application, the user equipment can include a processor, a memory and a transceiver unit which can include a receiver and a transmitter, the processor, the memory, the receiver and the transmitter can be connected through a bus system, the memory is used to store computer instructions, and the processor is used to execute the computer instructions stored in the memory to control the transceiver unit to transceive signals.

[0099] As an implementation manner, the functions of the receiver and the transmitter in the present application can be realized by a transceiver circuit or a dedicated chip for transceiving, and the processor can be realized by a dedicated processing chip, a processing circuit or a general-purpose chip.

[0100] As another implementation manner, the server provided by the embodiments of the present application can be implemented by using a general computer. That is, program codes for implementing the functions of the processor, the receiver and the transmitter are stored in the memory, and the general processor implements the functions of the processor, the receiver and the transmitter by executing the codes in the memory.

[0101] The embodiments of the present application further provide a computer readable storage medium, which stores a computer program. The computer program is executed by a processor to implement the steps of the fraud transaction account detection method. The computer readable storage medium can be a tangible storage medium, such as a random access memory (RAM), a memory, a read only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a floppy disk, a hard disk, a removable memory disk, a CD-ROM, or any other form of storage medium known in the art.

[0102] Those skilled in the art should understand that the exemplary components, systems and methods described in connection with the embodiments disclosed herein can be implemented in hardware, software or a combination thereof. The actual implementation depends on the specific application and design constraints imposed on the overall system. Those skilled in the art can use various methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application. When implemented in hardware, it can be, for example, an electronic circuit, an application specific integrated circuit (ASIC), appropriate firmware, a plug-in, a functional card, etc. When implemented in software, the elements of the present application are program or code segments used to perform the required tasks. The program or code segments can be stored in a machine readable medium or transmitted through a data signal carried in a carrier wave in a transmission medium or a communication link.

[0103] It should be noted that the present application is not limited to the specific configurations and processes described above and shown in the drawings. For the sake of brevity, detailed descriptions of well-known methods are omitted. In the above embodiments, several specific steps are described and shown as examples. However, the method processes of the present application are not limited to the specific steps described and shown, and those skilled in the art can make various changes, modifications and additions, or change the order of the steps, after understanding the spirit of the present application.

[0104] In the present application, the features described and / or illustrated for one embodiment can be used in the same way or in a similar way in one or more other embodiments, and / or in combination with or instead of features of other embodiments.

[0105] The above descriptions are only the preferred embodiments of the present application, and are not intended to limit the present application. The embodiments of the present application can be variously changed and modified by those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the scope of protection of the present application.

Claims

1. A fraud transaction account detection method based on a heterogeneous graph convolution network, characterized in that, The method comprises the following steps: A bipartite graph is constructed by taking a plurality of target transaction accounts as account nodes, taking a plurality of orders associated with each target transaction account as order nodes, and taking the association between each target transaction account and each order associated therewith as an order edge, wherein the order edge comprises an order submission edge and an order cancellation edge; The bipartite graph is convoluted based on a heterogeneous graph convolution network to obtain an account node hidden state corresponding to the account node, an order submission edge hidden state, and an order cancellation edge hidden state; The account node hidden state, the order submission edge hidden state, and the order cancellation edge hidden state are input into a pre-acquired classification model to enable the classification model to output an account anomaly detection result data of the account node; The target transaction account is a carbon transaction account, and the order is a carbon transaction order; The method comprises the following steps: Each order node hidden state corresponding to each order node is obtained based on the heterogeneous graph convolution network; Each order edge hidden state corresponding to each order edge is obtained based on the heterogeneous graph convolution network; The order submission edge hidden state is obtained based on each order node hidden state, each order edge hidden state, a combination function, a first aggregation function, and an order submission edge training weight matrix; The order cancellation edge hidden state is obtained based on each order node hidden state, each order edge hidden state, the combination function, a second aggregation function, and an order cancellation edge training weight matrix; The account node hidden state is obtained based on the order submission edge hidden state and the order cancellation edge hidden state of the previous layer, an account node hidden state, the combination function, an account node training weight matrix, and a third aggregation function; The order submission edge hidden state is obtained based on each order node hidden state, each order edge hidden state, a combination function, a first aggregation function, and an order submission edge training weight matrix, and the corresponding formula is as follows: wherein, denotes an order submission edge hidden state, denotes a nonlinear activation function, denotes an order submission edge training weight matrix, denotes a first aggregation function, denotes an order node hidden state, denotes an order edge hidden state, denotes an account node, denotes an order node, and concat denotes a concatenation function. The order cancellation edge hidden state is obtained based on each order node hidden state, each order edge hidden state, the combination function, a second aggregation function, and an order cancellation edge training weight matrix, and the corresponding formula is as follows: wherein, denotes an order cancel edge hidden state, denotes a nonlinear activation function, denotes an order cancel edge training weight matrix, denotes a second aggregation function, denotes an order node hidden state, denotes an order edge hidden state, denotes an account node, denotes an order node, concat denotes a concatenation function; The account node hidden state is obtained based on the order submission edge hidden state and the order cancellation edge hidden state of the previous layer, an account node hidden state, the combination function, an account node training weight matrix, and a third aggregation function, and the corresponding formula is as follows: wherein, denotes an account node hidden state, denotes a non-linear activation function, denotes an account node training weight matrix, denotes a third aggregation function, denotes a previous layer account node hidden state, denotes a previous layer order submission edge hidden state, denotes a previous layer order cancellation edge hidden state, and concat denotes a concatenation function. 2.The fraud transaction account detection method based on heterogeneous graph convolutional network according to claim 1, characterized in that, Each order node hidden state corresponding to each order node is obtained based on the account node hidden state, the order node hidden state, the order edge hidden state, and the order node training weight matrix of the previous layer corresponding to each order node. Each order node hidden state corresponding to each order node is obtained based on the account node hidden state, the order node hidden state, the order edge hidden state, and the order node training weight matrix of the previous layer corresponding to each order node. 3.The fraud transaction account detection method based on heterogeneous graph convolutional network according to claim 1, characterized in that, The order edge hidden state corresponding to each of the order edges is obtained by performing convolution processing on each of the order edges based on the heterogeneous graph convolution network, and includes: The order edge hidden state corresponding to each of the order edges is obtained based on the account node hidden state, the order node hidden state, the order edge hidden state of the previous layer, and the order edge training weight matrix corresponding to each of the order edges.

4. A computer device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor executes the computer program to implement the fraud transaction account detection method based on the heterogeneous graph convolution network as claimed in any one of claims 1 to 3.

5. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the fraud transaction account detection method based on the heterogeneous graph convolution network as claimed in any one of claims 1 to 3.

6. A computer program product comprising a computer program, characterized in that, The computer program is executed by the processor to implement the fraud transaction account detection method based on the heterogeneous graph convolution network as claimed in any one of claims 1 to 3.

Images