Target transaction identification method and device, electronic equipment, medium and program product

Abstract

The invention provides a target transaction identification method and device based on a knowledge graph, electronic equipment, a medium and a computer program product. The method and the device can be used in the technical field of artificial intelligence and information security. The method comprises the steps that a pre-constructed knowledge graph is input into a pre-trained target recognition probability model, the recognition probability of each transaction node in the knowledge graph is obtained, the knowledge graph comprises transaction nodes, intention nodes, transaction edges and intention edges, each transaction node in the knowledge graph is traversed in the target recognition probability model, and the recognition probability of each transaction node in the knowledge graph is obtained; the recognition probability of the traversal transaction node is calculated according to a first neighbor node and a second neighbor node of the traversal transaction node until the recognition probability of each transaction node in the knowledge graph is calculated, and the first neighbor node is an intention node having an intention edge with the traversal transaction node; the second neighbor node is a transaction node having a transaction edge with the traversal transaction node; and determining a target transaction meeting a preset condition according to the identification probability.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence and information security technology, and more specifically, to a target transaction identification method, device, electronic device, medium and computer program product based on a knowledge graph. Background Art

[0002] With the development of digital payment and digital currency, traditional monitoring systems based on rules and statistical methods have been unable to effectively deal with increasingly complex patterns that damage the property safety of users and secret abnormal behaviors. The existing technology predicts the possibility of abnormal transactions based on transaction information between users, but the accuracy of predicting abnormal transactions based on transaction information alone is low, which can easily cause normal transaction behaviors to be accidentally damaged, leading to customer complaints. Therefore, how to improve the accuracy of abnormal behavior prediction to achieve accurate crackdown on abnormal behaviors has become a problem that technicians in this field need to solve urgently. Summary of the invention

[0003] In view of this, the present disclosure provides a target transaction identification method, device, electronic device, computer-readable storage medium and computer program product based on a knowledge graph with high accuracy in identifying abnormal behaviors.

[0004] One aspect of the present disclosure provides a target transaction identification method based on a knowledge graph, wherein the knowledge graph includes transaction nodes and intention nodes, and the knowledge graph also includes transaction edges between the transaction nodes and the transaction nodes, and intention edges between the transaction nodes and the intention nodes. The method includes: inputting the pre-constructed knowledge graph into a pre-trained target recognition probability model to obtain the recognition probability of each transaction node in the knowledge graph, wherein the target recognition probability model is a neural network model based on an attention mechanism, traversing each of the transaction nodes in the knowledge graph in the target recognition probability model, calculating the recognition probability of the traversed transaction node according to the first neighbor node and the second neighbor node of the traversed transaction node, until the recognition probability of each of the transaction nodes in the knowledge graph is calculated, wherein the first neighbor node is the intention node having an intention edge with the traversed transaction node, and the second neighbor node is the transaction node having a transaction edge with the traversed transaction node; and determining the target transaction that meets the preset conditions according to the recognition probability.

[0005] According to the target transaction identification method based on the knowledge graph of the embodiment of the present disclosure, the identification probability of each transaction node in the knowledge graph is obtained by inputting a pre-constructed knowledge graph into a pre-trained target recognition probability model, wherein the knowledge graph includes transaction nodes and intention nodes, and the knowledge graph also includes transaction edges between transaction nodes and intention edges between transaction nodes and intention nodes. The target recognition probability model is a neural network model based on the attention mechanism. Each transaction node in the knowledge graph is traversed in the target recognition probability model, and the identification probability of the traversed transaction node is calculated according to the first neighbor node and the second neighbor node of the traversed transaction node, until the identification probability of each transaction node in the knowledge graph is calculated, the first neighbor node is the intention node having an intention edge with the traversed transaction node, and the second neighbor node is the transaction node having a transaction edge with the traversed transaction node; the target transaction that meets the preset conditions is determined according to the identification probability. Since the knowledge graph disclosed in the present invention includes intention nodes and transaction nodes, when processing the knowledge graph, the target recognition probability model can propagate the information of the intention nodes adjacent to the traversed transaction nodes to the transaction nodes adjacent to the traversed transaction nodes, thereby converting a heterogeneous graph into a homogeneous graph, and improving the accuracy of abnormal transaction identification, thereby accurately combating abnormal behaviors and reducing complaints from normal customers.

[0006] In some embodiments, the target recognition probability model includes a node embedding model, an intention propagation model, an influence model and a probability calculation model, and the step of inputting a pre-constructed knowledge graph into a pre-trained target recognition probability model to obtain the recognition probability of each of the transaction nodes in the knowledge graph includes: inputting all the first neighbor nodes of the traversed transaction node into the node embedding model to obtain the intention vector representation corresponding to each of the first neighbor nodes; inputting the intention vector representation of all the first neighbor nodes of the traversed transaction node and all the second neighbor nodes of the traversed transaction node into the intention propagation model to obtain the intention vector representation of each of the second neighbor nodes that integrates all the first neighbor nodes; The transaction vector representation of the intention vector representation of the first neighbor node; the transaction vector representation of all the second neighbor nodes of the traversed transaction node are input into the influence model to obtain the vector representation of the traversed transaction node; the vector representation of the traversed transaction node is input into the probability calculation model to obtain the recognition probability of the traversed transaction node; and each of the transaction nodes in the knowledge graph is used as the traversed transaction node, and all the first neighbor nodes of the traversed transaction node are input into the node embedding model to obtain the intention vector representation corresponding to each first neighbor node, until the recognition probability of each of the transaction nodes in the knowledge graph is obtained.

[0007] In some embodiments, the step of inputting all first neighbor nodes of the traversed transaction node into the node embedding model to obtain the intention vector representation corresponding to each first neighbor node includes: inputting all first neighbor nodes of the traversed transaction node into the node embedding model, and using the Node2Vec algorithm in the node embedding model to generate the intention vector representation corresponding to each first neighbor node.

[0008] In some embodiments, the step of inputting the intention vector representations of all the first neighbor nodes of the traversed transaction node and all the second neighbor nodes of the traversed transaction node into the intention propagation model to obtain a transaction vector representation of each second neighbor node that integrates the intention vector representations of all the first neighbor nodes includes: inputting the intention vector representations of all the first neighbor nodes of the traversed transaction node and all the second neighbor nodes of the traversed transaction node into the intention propagation model; using an attention mechanism to assign a weight for the target second neighbor node to the intention vector representation of each first neighbor node; multiplying the weight of the intention vector representation of each first neighbor node of the target second neighbor node by the corresponding intention vector representation to obtain an intermediate vector representation; summing up the intermediate vector representations corresponding to all the first neighbor nodes to obtain the transaction vector representation of the target second neighbor node; and taking each second neighbor node among all the second neighbor nodes as the target second neighbor node, repeatedly executing the step of assigning a weight for the target second neighbor node to the intention vector representation of each first neighbor node using an attention mechanism until a transaction vector representation of each second neighbor node that integrates the intention vector representations of all the first neighbor nodes is obtained.

[0009] In some embodiments, the step of inputting the transaction vector representation of all the second neighbor nodes of the traversed transaction node into the influence model to obtain the vector representation of the traversed transaction node includes: inputting the transaction vector representation of all the second neighbor nodes of the traversed transaction node into the influence model, and using the graph attention neural network in the influence model to obtain the vector representation of the traversed transaction node.

[0010] In some embodiments, the step of inputting the vector representation of the traversed transaction node into the probability calculation model to obtain the recognition probability of the traversed transaction node includes: inputting the vector representation of the traversed transaction node into the probability calculation model, and using the fully connected layer in the probability calculation model to obtain the hidden vector representation of the traversed transaction node; normalizing the hidden vector representation to obtain the recognition probability of the traversed transaction node.

[0011] In some embodiments, the step of pre-building the knowledge graph includes: obtaining authorization from users in the system for transaction information of m transactions in the system and user behavior information within a time period of t seconds before each transaction, where m is an integer greater than or equal to 1, and t is greater than 0; after obtaining authorization from the user for the transaction information and the behavior information, determining the n transactions of the transaction according to the user behavior information within a time period of t seconds before each transaction. i Intent information, where n i is an integer greater than or equal to 1, i is an integer greater than or equal to 1 and less than or equal to m; and according to the transaction information of the m transactions and the n of each transaction in the m transactions i The knowledge graph is constructed based on intention information, wherein the transaction node and the transaction edge are constructed based on the transaction information, the intention node is constructed based on the intention information, and the intention edge is constructed based on the relationship between the transaction and the intention information.

[0012] Another aspect of the present disclosure provides a target transaction identification device based on a knowledge graph, wherein the knowledge graph includes transaction nodes and intention nodes, and the knowledge graph also includes transaction edges between the transaction nodes and the transaction nodes, and intention edges between the transaction nodes and the intention nodes. The device includes: a probability calculation module, wherein the probability calculation module is used to input the pre-constructed knowledge graph into a pre-trained target recognition probability model to obtain the recognition probability of each transaction node in the knowledge graph, wherein the target recognition probability model is a neural network model based on an attention mechanism, wherein each transaction node in the knowledge graph is traversed in the target recognition probability model, and the recognition probability of the traversed transaction node is calculated according to the first neighbor node and the second neighbor node of the traversed transaction node, until the recognition probability of each transaction node in the knowledge graph is calculated, wherein the first neighbor node is the intention node having an intention edge with the traversed transaction node, and the second neighbor node is the transaction node having a transaction edge with the traversed transaction node; and a determination module, wherein the determination module is used to determine the target transaction that meets the preset conditions according to the recognition probability.

[0013] Another aspect of the present disclosure provides an electronic device, including one or more processors and one or more memories, wherein the memories are used to store executable instructions, and when the executable instructions are executed by the processors, the above-mentioned method is implemented.

[0014] Another aspect of the present disclosure provides a computer-readable storage medium storing computer-executable instructions, which are used to implement the above method when executed.

[0015] Another aspect of the present disclosure provides a computer program product, comprising a computer program, wherein the computer program comprises computer executable instructions, and the instructions are used to implement the method as described above when executed. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The above and other objects, features and advantages of the present disclosure will become more apparent through the following description of the embodiments of the present disclosure with reference to the accompanying drawings, in which:

[0017] Figure 1 An exemplary system architecture to which the method and apparatus according to an embodiment of the present disclosure can be applied is schematically shown;

[0018] Figure 2 A flowchart of a target transaction identification method based on a knowledge graph according to an embodiment of the present disclosure is schematically shown;

[0019] Figure 3 A flowchart schematically shows the steps of inputting a pre-built knowledge graph into a pre-trained target recognition probability model to obtain the recognition probability of each transaction node in the knowledge graph according to an embodiment of the present disclosure;

[0020] Figure 4 A flowchart schematically shows the steps of inputting the intention vector representations of all first neighbor nodes of a transaction node and all second neighbor nodes of a transaction node into an intention propagation model to obtain a transaction vector representation of each second neighbor node that incorporates the intention vector representations of all first neighbor nodes according to an embodiment of the present disclosure;

[0021] Figure 5 A flowchart schematically shows the steps of inputting the vector representation of the traversed transaction node into the probability calculation model to obtain the recognition probability of the traversed transaction node according to an embodiment of the present disclosure;

[0022] Figure 6 A flowchart schematically shows the steps of pre-building a knowledge graph according to an embodiment of the present disclosure;

[0023] Figure 7 A schematic diagram of a knowledge graph according to an embodiment of the present disclosure is schematically shown;

[0024] Figure 8 A schematic diagram schematically shows a spliced ​​intent information list according to an embodiment of the present disclosure;

[0025] Fig. 9 Schematically shows a flowchart of intent node embedding according to an embodiment of the present disclosure;

[0026] Fig.10 A flowchart of message propagation of each intended neighbor node according to an embodiment of the present disclosure is schematically shown;

[0027] Fig.11 Schematically shows a flow chart of each transaction neighbor node entering GAT training according to an embodiment of the present disclosure;

[0028] Fig.12 A flowchart schematically shows the output probability of the normalized access to the fully connected layer according to an embodiment of the present disclosure;

[0029] Fig.13 A block diagram of a target transaction identification device based on a knowledge graph according to an embodiment of the present disclosure is schematically shown;

[0030] Fig.14 A block diagram of an electronic device according to an embodiment of the present disclosure is schematically shown. DETAILED DESCRIPTION

[0031] Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the present disclosure. In the following detailed description, for ease of explanation, many specific details are set forth to provide a comprehensive understanding of the embodiments of the present disclosure. However, it is apparent that one or more embodiments may also be implemented without these specific details. In addition, in the following description, descriptions of known structures and technologies are omitted to avoid unnecessary confusion of the concepts of the present disclosure.

[0032] In the technical solution of this disclosure, the acquisition, storage and application of user personal information involved are in compliance with the provisions of relevant laws and regulations, necessary confidentiality measures are taken, and they do not violate public order and good morals. In the technical solution of this disclosure, the acquisition, collection, storage, use, processing, transmission, provision, disclosure and application of data are in compliance with the provisions of relevant laws and regulations, necessary confidentiality measures are taken, and they do not violate public order and good morals.

[0033] The terms used herein are only for describing specific embodiments and are not intended to limit the present disclosure. The terms "comprise", "include", etc. used herein indicate the existence of the features, steps, operations and / or components, but do not exclude the existence or addition of one or more other features, steps, operations or components.

[0034] When using expressions such as "at least one of A, B or C, etc.", it should generally be interpreted in accordance with the meaning of the expression generally understood by those skilled in the art (for example, "a system having at least one of A, B or C" should include but is not limited to a system having A alone, B alone, C alone, A and B, A and C, B and C, and / or A, B, C, etc.). The terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, features defined as "first" and "second" may explicitly or implicitly include one or more of the said features.

[0035] With the development of digital payment and digital currency, traditional monitoring systems based on rules and statistical methods have been unable to effectively deal with increasingly complex patterns that damage the property safety of users and secret abnormal behaviors. The existing technology predicts the possibility of abnormal transactions based on transaction information between users, but the accuracy of predicting abnormal transactions based on transaction information alone is low, which can easily cause normal transaction behaviors to be accidentally damaged, leading to customer complaints. Therefore, how to improve the accuracy of abnormal behavior prediction to achieve accurate crackdown on abnormal behaviors has become a problem that technicians in this field need to solve urgently.

[0036] The embodiments of the present disclosure provide a method, device, electronic device, computer-readable storage medium and computer program product for target transaction identification based on a knowledge graph. The knowledge graph includes transaction nodes and intention nodes, and the knowledge graph also includes transaction edges between transaction nodes and intention edges between transaction nodes and intention nodes. The target transaction identification method based on the knowledge graph includes: inputting a pre-built knowledge graph into a pre-trained target recognition probability model to obtain the recognition probability of each transaction node in the knowledge graph, wherein the target recognition probability model is a neural network model based on an attention mechanism, traversing each transaction node in the knowledge graph in the target recognition probability model, calculating the recognition probability of the traversed transaction node according to the first neighbor node and the second neighbor node of the traversed transaction node, until the recognition probability of each transaction node in the knowledge graph is calculated, the first neighbor node is the intention node with an intention edge with the traversed transaction node, and the second neighbor node is the transaction node with a transaction edge with the traversed transaction node; and determining the target transaction that meets the preset conditions according to the recognition probability.

[0037] It should be noted that the knowledge graph-based target transaction identification method, device, electronic device, computer-readable storage medium and computer program product disclosed in the present invention can be used in the field of artificial intelligence technology, and can also be used in any field other than the field of artificial intelligence and information security technology, such as the financial field. The field of the present invention is not limited here.

[0038] Figure 1The exemplary system architecture 100 to which the knowledge graph-based target transaction identification method, apparatus, electronic device, computer-readable storage medium, and computer program product can be applied according to an embodiment of the present disclosure is schematically shown. It should be noted that: Figure 1 What is shown is merely an example of a system architecture to which the embodiments of the present disclosure can be applied, in order to help those skilled in the art understand the technical content of the present disclosure, but it does not mean that the embodiments of the present disclosure cannot be used in other devices, systems, environments or scenarios.

[0039] like Figure 1 As shown, the system architecture 100 according to this embodiment may include terminal devices 101, 102, 103, a network 104 and a server 105. The network 104 is used to provide a medium for communication links between the terminal devices 101, 102, 103 and the server 105. The network 104 may include various connection types, such as wired, wireless communication links or optical fiber cables, etc.

[0040] Users can use terminal devices 101, 102, 103 to interact with server 105 through network 104 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 101, 102, 103, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, social platform software, etc. (only examples).

[0041] The terminal devices 101 , 102 , and 103 may be various electronic devices having a display screen and supporting web browsing, including but not limited to smart phones, tablet computers, laptop computers, and desktop computers.

[0042] The server 105 may be a server that provides various services, such as a background management server (only an example) that provides support for websites browsed by users using the terminal devices 101, 102, and 103. The background management server may analyze and process the received data such as user requests, and feed back the processing results (such as web pages, information, or data obtained or generated according to user requests) to the terminal device.

[0043] It should be noted that the target transaction identification method based on the knowledge graph provided in the embodiment of the present disclosure can generally be executed by the server 105. Accordingly, the target transaction identification device based on the knowledge graph provided in the embodiment of the present disclosure can generally be set in the server 105. The target transaction identification method based on the knowledge graph provided in the embodiment of the present disclosure can also be executed by a server or server cluster that is different from the server 105 and can communicate with the terminal devices 101, 102, 103 and / or the server 105. Accordingly, the target transaction identification device based on the knowledge graph provided in the embodiment of the present disclosure can also be set in a server or server cluster that is different from the server 105 and can communicate with the terminal devices 101, 102, 103 and / or the server 105.

[0044] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0045] The following will be based on Figure 1 The scene described by Figure 2 to Figure 5 The target transaction identification method based on knowledge graph in an embodiment of the present disclosure is described in detail.

[0046] Figure 2 The flowchart of the target transaction identification method based on the knowledge graph according to an embodiment of the present disclosure is schematically shown.

[0047] like Figure 2 As shown, the target transaction identification method based on the knowledge graph of this embodiment includes operation S210 and operation S220.

[0048] In operation S210, the pre-built knowledge graph is input into the pre-trained target recognition probability model to obtain the recognition probability of each transaction node in the knowledge graph. The knowledge graph includes transaction nodes and intent nodes, and the knowledge graph also includes transaction edges between transaction nodes and intent edges between transaction nodes and intent nodes. The target recognition probability model is a neural network model based on the attention mechanism. In the target recognition probability model, each transaction node in the knowledge graph is traversed, and the recognition probability of the traversed transaction node is calculated according to the first neighbor node and the second neighbor node of the traversed transaction node, until the recognition probability of each transaction node in the knowledge graph is calculated, the first neighbor node is the intent node with the intent edge with the traversed transaction node, and the second neighbor node is the transaction node with the transaction edge with the traversed transaction node.

[0049] As a possible implementation method, the target recognition probability model includes a node embedding model, an intention propagation model, an influence model, and a probability calculation model, such as Figure 3As shown, operation S210 inputs the pre-built knowledge graph into the pre-trained target recognition probability model to obtain the recognition probability of each transaction node in the knowledge graph, including operations S211 to S215.

[0050] In operation S211, all first neighbor nodes of the traversed transaction node are input into the node embedding model to obtain the intention vector representation corresponding to each first neighbor node.

[0051] In some specific examples, operation S211 traverses all first neighbor nodes of the transaction node into the node embedding model to obtain the intention vector representation corresponding to each first neighbor node, including operation S2111.

[0052] In operation S2111, all first neighbor nodes of the traversed transaction node are input into the node embedding model, and the Node2Vec algorithm in the node embedding model is used to generate an intent vector representation corresponding to each first neighbor node.

[0053] It can be understood that the first neighbor node of the traversed transaction node is the intention node that has an intention edge with the traversed transaction node. As will be seen below, the intention node is constructed based on the intention information. In the t second time period before each transaction, the user can have multiple user behaviors, and n user behaviors can be extracted from the multiple user behaviors. i Intent information, each of which can construct an intent node, each of which can be a behavior sequence, is one-hot encoded, and then the Node2Vec algorithm is used to perform random walks according to the intent edge to form a fixed-length node sequence, and a low-dimensional vector representation of each intent node is generated. The low-dimensional vector representation can be used as the intent vector representation of each intent node. Operation S2111 can facilitate the step of inputting all first neighbor nodes of the traversed transaction node into the node embedding model to obtain the intent vector representation corresponding to each first neighbor node.

[0054] In operation S212, the intention vector representations of all first neighbor nodes traversing the transaction node and all second neighbor nodes traversing the transaction node are input into the intention propagation model to obtain a transaction vector representation of each second neighbor node that integrates the intention vector representations of all first neighbor nodes.

[0055] In some specific examples, such as Figure 4 As shown, operation S212 inputs the intention vector representations of all first neighbor nodes of the traversed transaction node and all second neighbor nodes of the traversed transaction node into the intention propagation model to obtain the transaction vector representation of each second neighbor node that integrates the intention vector representations of all first neighbor nodes, including operations S2121 to S2125.

[0056] In operation S2121, the intention vector representations of all first neighbor nodes of the traversed transaction node and all second neighbor nodes of the traversed transaction node are input into the intention propagation model.

[0057] In operation S2122, the attention mechanism is used to assign a weight to the target second neighbor node for the intention vector representation of each first neighbor node, wherein the target second neighbor node is any second neighbor node among all second neighbor nodes traversing the transaction node.

[0058] In operation S2123, the weight of the intention vector representation of each first neighbor node of the target second neighbor node is multiplied by the corresponding intention vector representation to obtain an intermediate vector representation.

[0059] In operation S2124, the intermediate vector representations corresponding to all first neighbor nodes are summed to obtain a transaction vector representation of the target second neighbor node.

[0060] In operation S2125, each second neighbor node among all the second neighbor nodes is taken as the target second neighbor node, and the step of using the attention mechanism to assign a weight to the target second neighbor node to the intention vector representation of each first neighbor node is repeatedly performed until a transaction vector representation of each second neighbor node is obtained that integrates the intention vector representations of all the first neighbor nodes.

[0061] It can be understood that the intention vector representation of each first neighbor node obtained by operation S2111 is arranged according to the time sequence of the behavior, connected to the long short-term memory neural network (LSTM), the timing information between the behavior sequences is obtained, and the hidden state of each first neighbor node is obtained. Then the attention mechanism is added to train the weight of each first neighbor node for the target second neighbor node, and the weight of the intention vector representation of each first neighbor node for the target second neighbor node is multiplied by the corresponding intention vector representation to obtain the intermediate vector representation. The transaction vector representation of the target second neighbor node can be obtained by summing the intermediate vector representations corresponding to all first neighbor nodes. At this time, the transaction vector representation of the target second neighbor node has integrated the embedded representation of the adjacent intention nodes of the traversed transaction node, and the intention node information is propagated to the transaction nodes adjacent to the traversed transaction node. Compared with the graph attention neural network directly using the intention and transaction nodes for probability calculation, the heterogeneous graph is converted into a homogeneous graph, which can be directly connected to the graph attention neural network for probability calculation.

[0062] Operations S2121 to S2125 can facilitate the input of the intention vector representations of all first neighbor nodes traversing the transaction node and all second neighbor nodes traversing the transaction node into the intention propagation model to obtain the transaction vector representation of each second neighbor node that integrates the intention vector representations of all first neighbor nodes.

[0063] In operation S213, the transaction vector representations of all second neighbor nodes of the traversed transaction node are input into the influence model to obtain the vector representation of the traversed transaction node.

[0064] In some specific examples, operation S213 inputs the transaction vector representation of all second neighbor nodes of the traversed transaction node into the influence model to obtain the vector representation of the traversed transaction node, including operation S2131.

[0065] In operation S2131, the transaction vector representations of all second neighbor nodes of the traversed transaction node are input into the influence model, and the graph attention neural network in the influence model is used to obtain the vector representation of the traversed transaction node. The transaction nodes adjacent to the traversed transaction node are connected to the graph attention neural network (GAT). Different neighbor transaction nodes have different attention coefficients (i.e., weights) for the transaction node currently being traversed, thereby increasing the influence weight of abnormal transactions and reducing the weight of daily or disguised transactions, and obtaining the final vector representation of the traversed transaction node. Operation S2131 can facilitate the input of the transaction vector representations of all second neighbor nodes of the traversed transaction node into the influence model to obtain the vector representation of the traversed transaction node.

[0066] In operation S214, the vector representation of the traversed transaction node is input into the probability calculation model to obtain the recognition probability of the traversed transaction node. According to operations S211 to S214, on the basis of the homogeneous graph neural network based on the transaction node, the heterogeneous intention nodes are integrated, and the abnormal transactions and ordinary transactions have different influence weights, which can effectively identify the abnormal transactions disguised as normal transactions; at the same time, judging based on the transaction dimension can effectively reduce the impact of misblocking and malicious complaints.

[0067] In some specific examples, such as Figure 5 As shown, operation S214 inputs the vector representation of the traversed transaction node into the probability calculation model to obtain the recognition probability of the traversed transaction node, including operation S2141 and operation S2142.

[0068] In operation S2141, the vector representation of the traversed transaction node is input into the probability calculation model, and the hidden vector representation of the traversed transaction node is obtained by using the fully connected layer in the probability calculation model.

[0069] In operation S2142, the hidden vector representation is normalized to obtain the recognition probability of the traversed transaction node. The hidden vector representation can be normalized by a sigmoid function. Operations S2141 and S2142 can facilitate the input of the vector representation of the traversed transaction node into the probability calculation model to obtain the recognition probability of the traversed transaction node.

[0070] In operation S215, each transaction node in the knowledge graph is used as a traversed transaction node, and the step of inputting all first neighbor nodes of the traversed transaction node into the node embedding model to obtain the intention vector representation corresponding to each first neighbor node is repeatedly performed until the recognition probability of each transaction node in the knowledge graph is obtained.

[0071] Operations S211 to S215 can facilitate the input of the pre-built knowledge graph into the pre-trained target recognition probability model to obtain the recognition probability of each transaction node in the knowledge graph.

[0072] In operation S220, a target transaction that meets a preset condition is determined according to the recognition probability.

[0073] According to the target transaction identification method based on the knowledge graph of the embodiment of the present disclosure, the identification probability of each transaction node in the knowledge graph is obtained by inputting a pre-constructed knowledge graph into a pre-trained target recognition probability model, wherein the knowledge graph includes transaction nodes and intention nodes, and the knowledge graph also includes transaction edges between transaction nodes and intention edges between transaction nodes and intention nodes. The target recognition probability model is a neural network model based on the attention mechanism. Each transaction node in the knowledge graph is traversed in the target recognition probability model, and the identification probability of the traversed transaction node is calculated according to the first neighbor node and the second neighbor node of the traversed transaction node, until the identification probability of each transaction node in the knowledge graph is calculated, the first neighbor node is the intention node having an intention edge with the traversed transaction node, and the second neighbor node is the transaction node having a transaction edge with the traversed transaction node; the target transaction that meets the preset conditions is determined according to the identification probability. Since the knowledge graph disclosed in the present invention includes intention nodes and transaction nodes, when processing the knowledge graph, the target recognition probability model can propagate the information of the intention nodes adjacent to the traversed transaction nodes to the transaction nodes adjacent to the traversed transaction nodes, thereby converting a heterogeneous graph into a homogeneous graph, and improving the accuracy of abnormal transaction identification, thereby accurately combating abnormal behaviors and reducing complaints from normal customers.

[0074] Figure 6 A flowchart schematically illustrates the steps of pre-building a knowledge graph according to an embodiment of the present disclosure.

[0075] like Figure 6 As shown, the steps of pre-building a knowledge graph include operations S310 to S330.

[0076] In operation S310, authorization of users in the system for transaction information of m transactions in the system and user behavior information in a time period of t seconds before each transaction is obtained, where m is an integer greater than or equal to 1, and t is greater than 0.

[0077] In operation S320, after obtaining the user's authorization for the transaction information and behavior information, the transaction time n is determined according to the user's behavior information in the t seconds before each transaction. i Intent information, where n i is an integer greater than or equal to 1, and i is an integer greater than or equal to 1 and less than or equal to m.

[0078] In operation S330, based on the transaction information of the m transactions and the n of each transaction in the m transactions i The knowledge graph is constructed based on the intention information, where transaction nodes and transaction edges are constructed based on the transaction information, intent nodes are constructed based on the intention information, and intent edges are constructed based on the relationship between transactions and intent information.

[0079] It is understandable that transaction information may include but is not limited to specific transactions such as transfers and purchases; user behavior information in the t-second period before each transaction may include but is not limited to user behaviors such as searches, clicks, account inquiries, and homepages; a user may have multiple user behaviors in the t-second period before each transaction, and n user behaviors may be extracted from multiple user behaviors. i Each piece of intent information can construct an intent node, and each piece of intent information can be a behavior sequence.

[0080] like Figure 7 As shown in Figure 1, the nodes of the knowledge graph can be divided into two categories: transaction nodes and intention nodes. Adaptively, transaction nodes are specific transactions such as transfers and purchases, and intention nodes are obtained from user behavior sequences. We sort out various client tracking logs, record user behavior one hour before the transaction, perform label encoding on each action, and splice them into a list in order. The spliced ​​intent information list can be as follows: Figure 8 As shown. The edges of the knowledge graph are also divided into two categories: transaction edges between transaction nodes and intention edges between intention nodes. The principle of establishing transaction edges is: having the same device, IP address, transaction object, transaction amount, transaction product and / or remarks. The principle of establishing intention edges is: establishing intention edges between the intention node constructed based on the intention information extracted from the user behavior obtained one hour before the transaction and the transaction node corresponding to the transaction.

[0081] Operations S310 to S330 can facilitate the pre-construction of a knowledge graph.

[0082] Refer to the following Figure 9-12 The target transaction identification method based on the knowledge graph according to the embodiment of the present disclosure is described in detail. It is worth noting that the following description is only an exemplary description and is not a specific limitation of the present disclosure.

[0083] The disclosed process can be divided into two major steps. The first step is to construct a graph, taking transactions and user intentions as nodes, constructing transaction edges with the same devices, transaction users and amount purposes between transactions, and constructing intention edges with the user operation intentions of different clients within one hour before the transaction. The second step is to build the graph neural network model structure. First, DeepWalk is used to embed user intention nodes, and then the LSTM+attention mechanism is used to train the weights of the impact of different operation intentions on transaction nodes to obtain a new transaction node embedding expression. After that, all related transaction nodes are connected to the graph attention neural network, and the final embedding representation of the transaction node is obtained by training. Finally, the fully connected layer is connected for normalization to obtain the predicted abnormal probability of the transaction node. The present disclosure uses transactions as the judgment dimension, and embeds user intentions and related transactions together to represent transaction nodes. It can release ordinary transactions, intercept abnormal transactions of accounts, and block abnormal behaviors without banning accounts.

[0084] First, we build the graph. Nodes are divided into two categories: transaction nodes and intention nodes. Transaction nodes are specific transactions such as transfers and purchases, and intention nodes are obtained from user behavior sequences. We sort out various client tracking logs, record user behavior one hour before the transaction, perform label encoding on each action, and splice them into a list in order.

[0085] Edges are also divided into two categories: transaction edges between transaction nodes and intention edges between transaction nodes. The principle of establishing transaction edges is that if they have the same device, IP, transaction object, or the transaction amount and transaction product / remarks are the same, then edges are established between transaction nodes. The principle of establishing intention edges is that if the user has intention nodes on different clients one hour before the transaction occurs, an edge is established with the transaction.

[0086] Next, we will build the graph neural network structure. The first step is to Fig. 9 As shown in Figure 1, an embedded representation of the intent node is generated. Each operation sequence (user intent) is one-hot encoded, and then a DeepWalk random walk is performed according to the intent edge to form a fixed-length node sequence. The Skip-Gram model is used to train to obtain a low-dimensional vector representation h for each user intent node. I The second step is as follows: Fig.10 As shown in Figure 2, the intention node information is propagated to the adjacent transaction nodes. The trained user intention vector H I ={h I 1,h I 2,......,h I v}, where v is the number of intention node neighbors of the transaction node, which are arranged according to the time sequence of the behavior, connected to LSTM, and the time sequence information between the behavior sequences is obtained to obtain the hidden state of the intention node, and then connected to the training to obtain the weights α of different intention nodes i , we can get the transaction node representation h T , at this time h T The embedded representations of adjacent intent nodes have been fused. Compared with directly using intent and transaction nodes for graph neural network training, the heterogeneous graph is converted into a homogeneous graph and can be directly connected to the graph attention neural network training. Fig.11 As shown in the figure, adjacent transaction nodes are connected to the graph attention neural network (GAT) for training. Different neighbor nodes have different attention coefficients (i.e., weights) for the current node, which can increase the influence weight of abnormal transactions and reduce the weight of daily or disguised transactions, and obtain the final transaction node expression. The fourth step is Fig.12 As shown in the figure, it is finally connected to the fully connected layer and then normalized to obtain the recognition probability P of each transaction.

[0087] After obtaining the probability, the interception threshold can be set according to the data distribution to identify anomalies in the transaction dimension.

[0088] This paper proposes a method for identifying abnormal transactions that integrates user intentions. Using this method, on the basis of the traditional homogeneous graph neural network based on transaction nodes, heterogeneous intention nodes are integrated. Abnormal transactions and ordinary transactions have different influence weights, which can effectively identify abnormal transactions disguised as normal transactions; at the same time, judging based on transactions can effectively reduce the impact of misblocking and malicious complaints.

[0089] Based on the above-mentioned target transaction identification method based on knowledge graph, the present disclosure also provides a target transaction identification device based on knowledge graph. Fig.13 A target transaction identification device based on knowledge graph is described in detail.

[0090] Fig.13 The structural block diagram of a target transaction identification device based on a knowledge graph according to an embodiment of the present disclosure is schematically shown.

[0091] The target transaction identification device 10 based on knowledge graph includes a probability calculation module 1 and a determination module 2.

[0092] The probability calculation module 1 is used to perform operation S210: input the pre-constructed knowledge graph into the pre-trained target recognition probability model to obtain the recognition probability of each transaction node in the knowledge graph, wherein the knowledge graph includes transaction nodes and intent nodes, and the knowledge graph also includes transaction edges between transaction nodes and intent edges between transaction nodes and intent nodes. The target recognition probability model is a neural network model based on the attention mechanism. Each transaction node in the knowledge graph is traversed in the target recognition probability model, and the recognition probability of the traversed transaction node is calculated according to the first neighbor node and the second neighbor node of the traversed transaction node, until the recognition probability of each transaction node in the knowledge graph is calculated, the first neighbor node is the intent node having an intent edge with the traversed transaction node, and the second neighbor node is the transaction node having a transaction edge with the traversed transaction node.

[0093] The determination module 2 is used to perform operation S220: determine a target transaction that meets a preset condition according to the recognition probability.

[0094] According to some embodiments of the present disclosure, the target recognition probability model includes a node embedding model, an intention propagation model, an influence model and a probability calculation model, and the probability calculation module includes a first calculation unit, a second calculation unit, a third calculation unit, a fourth calculation unit and a repeated execution unit.

[0095] The first computing unit is used to input all first neighbor nodes of the traversed transaction node into the node embedding model to obtain the intention vector representation corresponding to each first neighbor node.

[0096] The second computing unit is used to input the intention vector representations of all first neighbor nodes traversing the transaction node and all second neighbor nodes traversing the transaction node into the intention propagation model to obtain a transaction vector representation of each second neighbor node that integrates the intention vector representations of all first neighbor nodes.

[0097] The third calculation unit is used to input the transaction vector representation of all second neighbor nodes of the traversed transaction node into the influence model to obtain the vector representation of the traversed transaction node.

[0098] The fourth calculation unit is used to input the vector representation of the traversed transaction node into the probability calculation model to obtain the recognition probability of the traversed transaction node.

[0099] The repeated execution unit is used to take each transaction node in the knowledge graph as a traversed transaction node, repeatedly execute the step of inputting all the first neighbor nodes of the traversed transaction node into the node embedding model, and obtaining the intention vector representation corresponding to each first neighbor node, until the recognition probability of each transaction node in the knowledge graph is obtained.

[0100] According to some embodiments of the present disclosure, the first computing unit includes a generating element.

[0101] The generation component is used to input all first neighbor nodes of the traversed transaction node into the node embedding model, and use the Node2Vec algorithm in the node embedding model to generate the intent vector representation corresponding to each first neighbor node.

[0102] According to some embodiments of the present disclosure, the second computing unit includes an input element, a distribution element, a first computing element, a second computing element, and a repeat execution element.

[0103] The input component is used to input the intention vector representations of all first neighbor nodes of the traversed transaction node and all second neighbor nodes of the traversed transaction node into the intention propagation model.

[0104] The allocation component is used to use the attention mechanism to allocate a weight to the target second neighbor node for the intention vector representation of each first neighbor node.

[0105] The first computing element is used to multiply the weight of the intention vector representation of each first neighbor node of the target second neighbor node by the corresponding intention vector representation to obtain an intermediate vector representation.

[0106] The second computing element is used to sum the intermediate vector representations corresponding to all first neighbor nodes to obtain the transaction vector representation of the target second neighbor node.

[0107] The repeated execution element is used to take each second neighbor node among all the second neighbor nodes as the target second neighbor node, and repeatedly execute the step of using the attention mechanism to assign a weight to the target second neighbor node to the intention vector representation of each first neighbor node, until a transaction vector representation of each second neighbor node is obtained that integrates the intention vector representations of all the first neighbor nodes.

[0108] According to some embodiments of the present disclosure, the third computing unit includes a third computing element.

[0109] The third computing element is used to input the transaction vector representation of all second neighbor nodes of the traversed transaction node into the influence model, and use the graph attention neural network in the influence model to obtain the vector representation of the traversed transaction node.

[0110] According to some embodiments of the present disclosure, the fourth computing unit includes a fourth computing element and a fifth computing element.

[0111] The fourth computing element is used to input the vector representation of the traversed transaction node into the probability computing model, and use the fully connected layer in the probability computing model to obtain the hidden vector representation of the traversed transaction node.

[0112] The fifth computing element is used to normalize the hidden vector representation to obtain the recognition probability of the traversed transaction node.

[0113] According to some embodiments of the present disclosure, the target transaction identification device based on the knowledge graph also includes a construction module, which is used to pre-construct the knowledge graph. The construction module includes an acquisition unit, a determination unit and a construction unit.

[0114] The acquisition unit is used to obtain the authorization of users in the system for transaction information of m transactions in the system and user behavior information in a time period of t seconds before each transaction, where m is an integer greater than or equal to 1 and t is greater than 0.

[0115] The determination unit is used to determine the transaction n according to the user behavior information in the t seconds before each transaction after obtaining the user's authorization for the transaction information and behavior information. i Intent information, where n i is an integer greater than or equal to 1, and i is an integer greater than or equal to 1 and less than or equal to m.

[0116] The construction unit is used to construct the transaction information of m transactions and n of each transaction in the m transactions. i The knowledge graph is constructed based on the intention information, where transaction nodes and transaction edges are constructed based on the transaction information, intent nodes are constructed based on the intention information, and intent edges are constructed based on the relationship between transactions and intent information.

[0117] According to the target transaction identification device 10 based on the knowledge graph of the embodiment of the present disclosure, the recognition probability of each transaction node in the knowledge graph is obtained by inputting the pre-constructed knowledge graph into the pre-trained target recognition probability model, wherein the knowledge graph includes transaction nodes and intention nodes, and the knowledge graph also includes transaction edges between transaction nodes and intention edges between transaction nodes and intention nodes. The target recognition probability model is a neural network model based on the attention mechanism. Each transaction node in the knowledge graph is traversed in the target recognition probability model, and the recognition probability of the traversed transaction node is calculated according to the first neighbor node and the second neighbor node of the traversed transaction node, until the recognition probability of each transaction node in the knowledge graph is calculated, the first neighbor node is the intention node having an intention edge with the traversed transaction node, and the second neighbor node is the transaction node having a transaction edge with the traversed transaction node; the target transaction that meets the preset conditions is determined according to the recognition probability. Since the knowledge graph disclosed in the present invention includes intention nodes and transaction nodes, when processing the knowledge graph, the target recognition probability model can propagate the information of the intention nodes adjacent to the traversed transaction nodes to the transaction nodes adjacent to the traversed transaction nodes, thereby converting a heterogeneous graph into a homogeneous graph, and improving the accuracy of abnormal transaction identification, thereby accurately combating abnormal behaviors and reducing complaints from normal customers.

[0118] In addition, according to an embodiment of the present disclosure, any multiple modules in the probability calculation module 1 and the determination module 2 can be combined into one module for implementation, or any one of the modules can be split into multiple modules. Alternatively, at least part of the functions of one or more of these modules can be combined with at least part of the functions of other modules and implemented in one module.

[0119] According to an embodiment of the present disclosure, at least one of the probability calculation module 1 and the determination module 2 can be at least partially implemented as a hardware circuit, such as a field programmable gate array (FPGA), a programmable logic array (PLA), a system on a chip, a system on a substrate, a system on a package, an application-specific integrated circuit (ASIC), or can be implemented by hardware or firmware such as any other reasonable way of integrating or packaging the circuit, or can be implemented in any one of the three implementation methods of software, hardware and firmware, or in an appropriate combination of any of them.

[0120] Alternatively, at least one of the probability calculation module 1 and the determination module 2 may be at least partially implemented as a computer program module, and when the computer program module is executed, the corresponding function may be executed.

[0121] Fig.14 A block diagram of an electronic device suitable for implementing the above method according to an embodiment of the present disclosure is schematically shown.

[0122] like Fig.14 As shown, the electronic device 900 according to an embodiment of the present disclosure includes a processor 901, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 902 or a program loaded from a storage part 908 into a random access memory (RAM) 903. The processor 901 may include, for example, a general-purpose microprocessor (e.g., a CPU), an instruction set processor and / or a related chipset and / or a special-purpose microprocessor (e.g., an application-specific integrated circuit (ASIC)), etc. The processor 901 may also include an onboard memory for caching purposes. The processor 901 may include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of the present disclosure.

[0123] In RAM 903, various programs and data required for the operation of electronic device 900 are stored. Processor 901, ROM 902 and RAM 903 are connected to each other via bus 904. Processor 901 performs various operations of the method flow according to the embodiment of the present disclosure by executing the program in ROM 902 and / or RAM 903. It should be noted that the program can also be stored in one or more memories other than ROM 902 and RAM 903. Processor 901 can also perform various operations of the method flow according to the embodiment of the present disclosure by executing the program stored in the one or more memories.

[0124] According to an embodiment of the present disclosure, the electronic device 900 may further include an input / output (I / O) interface 905, which is also connected to the bus 904. The electronic device 900 may further include one or more of the following components connected to the I / O interface 905: an input portion 906 including a keyboard, a mouse, etc.; an output portion 907 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage portion 908 including a hard disk, etc.; and a communication portion 909 including a network interface card such as a LAN card, a modem, etc. The communication portion 909 performs communication processing via a network such as the Internet. A drive 910 is also connected to the input / output (I / O) interface 905 as needed. A removable medium 911, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 910 as needed, so that a computer program read therefrom is installed into the storage portion 908 as needed.

[0125] The present disclosure also provides a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiments; or may exist independently without being assembled into the device / apparatus / system. The above computer-readable storage medium carries one or more programs, and when the above one or more programs are executed, the method according to the embodiment of the present disclosure is implemented.

[0126] According to an embodiment of the present disclosure, a computer-readable storage medium may be a non-volatile computer-readable storage medium, for example, may include but is not limited to: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program, which may be used by or in combination with an instruction execution system, an apparatus, or a device. For example, according to an embodiment of the present disclosure, a computer-readable storage medium may include the ROM 902 and / or RAM 903 described above and / or one or more memories other than ROM 902 and RAM 903.

[0127] The embodiment of the present disclosure also includes a computer program product, which includes a computer program, and the computer program contains program code for executing the method shown in the flowchart. When the computer program product is run in a computer system, the program code is used to enable the computer system to implement the method of the embodiment of the present disclosure.

[0128] The above functions defined in the system / device of the embodiment of the present disclosure are performed when the computer program is executed by the processor 901. According to the embodiment of the present disclosure, the system, device, module, unit, etc. described above can be implemented by a computer program module.

[0129] In one embodiment, the computer program may be based on a tangible storage medium such as an optical storage device, a magnetic storage device, etc. In another embodiment, the computer program may also be transmitted and distributed in the form of a signal on a network medium, and downloaded and installed through the communication part 909, and / or installed from a removable medium 911. The program code contained in the computer program may be transmitted using any appropriate network medium, including but not limited to: wireless, wired, etc., or any suitable combination of the above.

[0130] In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 909, and / or installed from the removable medium 911. When the computer program is executed by the processor 901, the above functions defined in the system of the embodiment of the present disclosure are performed. According to the embodiment of the present disclosure, the system, device, means, module, unit, etc. described above can be implemented by a computer program module.

[0131] According to an embodiment of the present disclosure, the program code for executing the computer program provided by the embodiment of the present disclosure can be written in any combination of one or more programming languages. Specifically, these computing programs can be implemented using high-level process and / or object-oriented programming languages, and / or assembly / machine languages. Programming languages ​​include, but are not limited to, Java, C++, python, "C" language or similar programming languages. The program code can be executed entirely on the user computing device, partially on the user device, partially on the remote computing device, or entirely on the remote computing device or server. In the case of a remote computing device, the remote computing device can be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computing device (for example, using an Internet service provider to connect through the Internet).

[0132] The flow charts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each box in the flow chart or block diagram can represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of a code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flow chart, and the combination of the boxes in the block diagram or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0133] It will be appreciated by those skilled in the art that the features described in the various embodiments and / or claims of the present disclosure may be combined and / or combined in a variety of ways, even if such combinations and / or combinations are not explicitly described in the present disclosure. In particular, the features described in the various embodiments and / or claims of the present disclosure may be combined and / or combined in a variety of ways without departing from the spirit and teachings of the present disclosure. All of these combinations and / or combinations fall within the scope of the present disclosure.

[0134] The embodiments of the present disclosure are described above. However, these embodiments are only for illustrative purposes and are not intended to limit the scope of the present disclosure. Although the embodiments are described above separately, this does not mean that the measures in the various embodiments cannot be used in combination to advantage. The scope of the present disclosure is defined by the attached claims and their equivalents. Without departing from the scope of the present disclosure, those skilled in the art may make a variety of substitutions and modifications, which should all fall within the scope of the present disclosure.

Claims

1. A target transaction identification method based on knowledge graph, characterized in that: The knowledge graph includes a transaction node and an intention node, the knowledge graph also includes a transaction edge between the transaction nodes and an intention edge between the transaction node and the intention node, and the method includes: Input the pre-constructed knowledge graph into a pre-trained target recognition probability model to obtain the recognition probability of each transaction node in the knowledge graph, wherein the target recognition probability model is a neural network model based on an attention mechanism, traverse each transaction node in the knowledge graph in the target recognition probability model, calculate the recognition probability of the traversed transaction node according to the first neighbor node and the second neighbor node of the traversed transaction node, until the recognition probability of each transaction node in the knowledge graph is calculated, the first neighbor node is the intention node having an intention edge with the traversed transaction node, and the second neighbor node is the transaction node having a transaction edge with the traversed transaction node; and A target transaction that meets a preset condition is determined according to the recognition probability.

2. The target transaction identification method based on knowledge graph according to claim 1 is characterized in that: The target recognition probability model includes a node embedding model, an intention propagation model, an influence model and a probability calculation model. The step of inputting the pre-built knowledge graph into the pre-trained target recognition probability model to obtain the recognition probability of each transaction node in the knowledge graph includes: Inputting all first neighbor nodes of the traversed transaction node into the node embedding model to obtain an intention vector representation corresponding to each first neighbor node; Inputting the intention vector representations of all the first neighboring nodes of the traversed transaction node and all the second neighboring nodes of the traversed transaction node into the intention propagation model to obtain a transaction vector representation of each second neighboring node that integrates the intention vector representations of all the first neighboring nodes; Inputting the transaction vector representations of all the second neighboring nodes of the traversed transaction node into the influence model to obtain the vector representation of the traversed transaction node; Inputting the vector representation of the traversed transaction node into the probability calculation model to obtain the recognition probability of the traversed transaction node; and Take each of the transaction nodes in the knowledge graph as the traversal transaction node, and repeatedly perform the step of inputting all the first neighbor nodes of the traversal transaction node into the node embedding model to obtain the intention vector representation corresponding to each of the first neighbor nodes, until the recognition probability of each of the transaction nodes in the knowledge graph is obtained.

3. The target transaction identification method based on knowledge graph according to claim 2 is characterized in that: The step of inputting all first neighbor nodes of the traversed transaction node into the node embedding model to obtain the intention vector representation corresponding to each first neighbor node includes: All first neighbor nodes of the traversed transaction node are input into the node embedding model, and the Node2Vec algorithm in the node embedding model is used to generate an intention vector representation corresponding to each first neighbor node.

4. The target transaction identification method based on knowledge graph according to claim 2 is characterized in that: The step of inputting the intention vector representations of all the first neighboring nodes of the traversed transaction node and all the second neighboring nodes of the traversed transaction node into the intention propagation model to obtain a transaction vector representation of each second neighboring node that integrates the intention vector representations of all the first neighboring nodes includes: Inputting the intention vector representations of all the first neighboring nodes of the traversed transaction node and all the second neighboring nodes of the traversed transaction node into the intention propagation model; Using an attention mechanism to assign a weight to a target second neighbor node to the intent vector representation of each of the first neighbor nodes; Multiplying the weight of the intention vector representation of each of the first neighbor nodes of the target second neighbor node by the corresponding intention vector representation to obtain an intermediate vector representation; Adding the intermediate vector representations corresponding to all the first neighbor nodes to obtain the transaction vector representation of the target second neighbor node; and Take each second neighbor node among all the second neighbor nodes as the target second neighbor node, and repeatedly perform the step of using the attention mechanism to assign a weight to the target second neighbor node for the intention vector representation of each first neighbor node, until a transaction vector representation of each second neighbor node is obtained that integrates the intention vector representations of all the first neighbor nodes.

5. The target transaction identification method based on knowledge graph according to claim 2 is characterized in that: The step of inputting the transaction vector representations of all the second neighboring nodes of the traversed transaction node into the influence model to obtain the vector representation of the traversed transaction node includes: The transaction vector representations of all the second neighbor nodes of the traversed transaction node are input into the influence model, and the graph attention neural network in the influence model is used to obtain the vector representation of the traversed transaction node.

6. The target transaction identification method based on knowledge graph according to claim 2 is characterized in that: The step of inputting the vector representation of the traversed transaction node into the probability calculation model to obtain the recognition probability of the traversed transaction node includes: Inputting the vector representation of the traversed transaction node into the probability calculation model, and using the fully connected layer in the probability calculation model to obtain the hidden vector representation of the traversed transaction node; The hidden vector representation is normalized to obtain the recognition probability of the traversed transaction node.

7. The target transaction identification method based on knowledge graph according to any one of claims 1 to 6, characterized in that: The steps of pre-building the knowledge graph include: Obtaining authorization of users in the system for transaction information of m transactions in the system and user behavior information within a time period of t seconds before each transaction, where m is an integer greater than or equal to 1 and t is greater than 0; After obtaining the user's authorization for the transaction information and the behavior information, the transaction is determined based on the user's behavior information within the t seconds before each transaction. i Intent information, where n i is an integer greater than or equal to 1, i is an integer greater than or equal to 1 and less than or equal to m; and According to the transaction information of the m transactions and the n of each transaction in the m transactions i The knowledge graph is constructed based on intention information, wherein the transaction node and the transaction edge are constructed based on the transaction information, the intention node is constructed based on the intention information, and the intention edge is constructed based on the relationship between the transaction and the intention information.

8. A target transaction identification device based on knowledge graph, characterized in that: The knowledge graph includes transaction nodes and intention nodes, the knowledge graph also includes transaction edges between the transaction nodes and intention edges between the transaction nodes and the intention nodes, and the device includes: a probability calculation module, wherein the probability calculation module is used to input the pre-constructed knowledge graph into a pre-trained target recognition probability model to obtain the recognition probability of each transaction node in the knowledge graph, wherein the target recognition probability model is a neural network model based on an attention mechanism, and each transaction node in the knowledge graph is traversed in the target recognition probability model, and the recognition probability of the traversed transaction node is calculated according to the first neighbor node and the second neighbor node of the traversed transaction node, until the recognition probability of each transaction node in the knowledge graph is calculated, wherein the first neighbor node is the intention node having an intention edge with the traversed transaction node, and the second neighbor node is the transaction node having a transaction edge with the traversed transaction node; and A determination module is used to execute a target transaction that satisfies a preset condition according to the recognition probability.

9. An electronic device, comprising: one or more processors; a storage device for storing one or more computer programs, It is characterized in that the one or more processors execute the one or more computer programs to implement the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

11. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images