Graph node relation representation generation and graph node service relation prediction method and device

Abstract

The embodiment of the invention provides a graph node relation representation generation method and device and a graph node service relation prediction method and device. When the graph node relation representation is generated, node representation propagation and node representation aggregation are executed from the first graph node and the second graph node respectively, and node representations of the second graph node and the first graph node are determined; and generating a node relationship characterization between the first and second graph nodes based on the node characterization of the first and second graph nodes. During node representation propagation, propagating a previous node propagation representation of each source graph node to each target graph node of the target graph node set of the source graph node; and generating a current node propagation representation of each target graph node according to the node propagation representation received by each target graph node and a previous node propagation representation of the target graph node. And during node characterization aggregation processing, according to the previous node characterization of the aggregation graph node and the previous node characterization of the neighbor graph node, generating the current node characterization of the aggregation graph node.

Description

[0001] This application is a divisional application of the invention patent application with application number 202111003074.7 filed on August 30, 2021, and entitled “Method and Device for Generating Graph Node Relationship Representation and Predicting Graph Node Business Relationship”. Technical Field

[0002] The embodiments of this specification generally relate to the field of graph data processing, and in particular to a graph node relationship representation generation method and apparatus for generating node relationship representations between graph nodes in graph data, and a graph node business relationship prediction method and apparatus for predicting business relationships between graph nodes in graph data. Background Art

[0003] In some business scenarios, business data is usually processed into graph data. In graph data, business entities such as users, enterprises and merchants are represented as nodes in the graph data, and the relationship between business entities is represented as the edge attribute relationship between nodes. When processing business, it is necessary to predict whether there is a business relationship between two given nodes (business entities) in the graph data. Then, subsequent business decision processing is performed based on the business relationship prediction results. For example, in the commodity recommendation scenario, graph data can be generated based on the historical commodity purchase data of the user's purchase of commodities, and the graph data can be used to predict whether there is a commodity purchase relationship between the user and a certain commodity, that is, whether the user will buy the commodity, and whether to recommend the commodity to the user based on the commodity purchase relationship prediction results. In the social scenario, graph data can be generated based on social data, and the graph data can be used to predict whether two users are friends, and subsequent social processing is performed based on the friend relationship prediction results. In the supply chain mining scenario, graph data can be generated based on the historical data of enterprise supply relationships, and the graph data can be used to predict whether two enterprises constitute a supply chain upstream and downstream relationship, and supply chain mining is performed based on the supply chain upstream and downstream relationship prediction results. Summary of the invention

[0004] In view of the above, the embodiments of this specification provide a method for generating a graph node relationship representation and a method and device for predicting a graph node business relationship. By using the method for generating a graph node relationship representation and the method and device for predicting a graph node business relationship, the obtained graph node relationship representation between graph nodes can accurately reflect the node attribute information of the graph nodes and the interaction structure information between the graph nodes, thereby improving the accuracy of predicting the business relationship between graph nodes.

[0005] According to one aspect of an embodiment of the present specification, a method for generating a node relationship representation between a first graph node and a second graph node in graph data is provided, wherein the graph data is generated based on business data, and the method comprises: performing node representation propagation and node representation aggregation starting from the first graph node and the second graph node, respectively, to determine the node representations of the second graph node and the first graph node; and generating a node relationship representation between the first graph node and the second graph node based on the node representations of the first graph node and the second graph node, wherein the node representation propagation and the node representation aggregation are performed for a specified number of iterations, and during the node representation propagation, the node of the previous iteration of each source graph node in the source graph node set of the current iteration is added. The propagation representation is propagated to each target graph node in the target graph node set of the source graph node, wherein the target graph node set is composed of the neighbor graph nodes of the source graph node; the node propagation representation of the current iteration process of each target graph node is generated according to the node propagation representation received by each target graph node and its own node propagation representation of the previous iteration process, and when the node representations are aggregated, the node representation of the current iteration process of the aggregate graph node is generated according to the node representation of the previous iteration process of the aggregate graph node and the node representation of the previous iteration process of the neighbor graph nodes of the aggregate graph node, wherein the aggregate graph node includes the first graph node or the second graph node, and the initial node representation of the graph node is generated based on the node propagation representation of the graph node and the original feature of the graph node.

[0006] Optionally, in an example of the above aspect, based on the node representations of the first graph node and the second graph node, generating a node relationship representation between the first graph node and the second graph node may include: splicing the node representations of the first graph node and the second graph node to generate a node relationship representation between the first graph node and the second graph node.

[0007] Optionally, in an example of the above aspect, generating the node propagation representation of the current iteration process of the target graph node based on the node propagation representations received by each target graph node and its own node propagation representation of the previous iteration process may include: generating the node propagation representation of the current iteration process of the target graph node based on the node propagation representations received by each target graph node from each neighbor graph node, the edge relationship characteristics between each neighbor graph node and its own node propagation representation of the previous iteration process.

[0008] Optionally, in an example of the above aspect, generating an initial node representation of a graph node based on a node propagation representation of the graph node and an original feature of the graph node may include: generating the initial node representation of the graph node by concatenating the node propagation representation of the graph node and the original feature of the graph node.

[0009] Optionally, in an example of the above aspect, the method is implemented based on a graph neural network.

[0010] Optionally, in an example of the above aspect, generating a node representation of a current iteration process of an aggregate graph node based on a node representation of a previous iteration process of the aggregate graph node and a node representation of a previous iteration process of a neighbor graph node of the aggregate graph node may include: using an aggregation function to aggregate the node representation of a previous iteration process of the aggregate graph node and the node representation of a previous iteration process of a neighbor graph node of the aggregate graph node to generate a node representation of a current iteration process of the aggregate graph node.

[0011] Optionally, in an example of the above aspect, the graph neural network includes a graph neural network with an Attention mechanism and an LSTM aggregator.

[0012] Optionally, in an example of the above aspect, generating the node propagation representation of the current iteration process of each target graph node based on the node propagation representation received by each target graph node and its own node propagation representation of the previous iteration process may include: aggregating the node propagation representation received by each target graph node through an Attention operation to obtain the neighbor node propagation representation of each target graph node; and performing LSTM aggregation on the neighbor node propagation representation of each target graph node and the node propagation representation of the previous iteration process of the target graph node through an LSTM aggregator to generate the node propagation representation of the current iteration process of each target graph node.

[0013] Optionally, in an example of the above aspect, the business data includes one of the following business data: social data, financial transaction data, commodity transaction data and enterprise supply relationship data.

[0014] According to another aspect of an embodiment of the present specification, a method for predicting a business relationship between a first graph node and a second graph node in graph data is provided, wherein the graph data is generated based on business data, and the method comprises: performing node representation propagation and node representation aggregation starting from the first graph node and the second graph node, respectively, to determine the node representations of the second graph node and the first graph node; generating a node relationship representation between the first graph node and the second graph node based on the node representations of the first graph node and the second graph node; and predicting the business relationship between the first graph node and the second graph node based on the node relationship representation between the first graph node and the second graph node, wherein the node representation propagation and the node representation aggregation are performed for a specified number of iterations, and when the node representation is propagated, the current iteration is repeated. The node propagation representation of the previous iteration process of each source graph node in the source graph node set is propagated to each target graph node in the target graph node set of the source graph node, and the target graph node set is composed of the neighbor graph nodes of the source graph node; the node propagation representation of the current iteration process of each target graph node is generated according to the node propagation representation received by each target graph node and its own node propagation representation of the previous iteration process, and when the node representations are aggregated, the node representation of the current iteration process of the aggregate graph node is generated according to the node representation of the previous iteration process of the aggregate graph node and the node representation of the previous iteration process of the neighbor graph nodes of the aggregate graph node, wherein the aggregate graph node includes the first graph node or the second graph node, and the initial node representation of the graph node is generated based on the node propagation representation of the graph node and the original feature of the graph node.

[0015] According to another aspect of an embodiment of the present specification, there is provided a device for generating a node relationship representation between a first graph node and a second graph node in graph data, wherein the graph data is generated based on business data, and the device comprises: a node representation determination unit, which starts node representation propagation and node representation aggregation from the first graph node and the second graph node, respectively, to determine the node representation of the second graph node and the first graph node; and a node relationship representation generation unit, which generates a node relationship representation between the first graph node and the second graph node based on the node representation of the first graph node and the second graph node, wherein the node representation determination unit comprises: a node representation propagation module, which propagates the node propagation representation of the previous iteration process of each source graph node in the source graph node set of the current iteration process to the target graph of the source graph node Each target graph node in a node set, wherein the target graph node set is composed of neighbor graph nodes of the source graph node, and a node propagation representation of a current iteration process of each target graph node is generated according to the node propagation representation received by each target graph node and its own node propagation representation of a previous iteration process; a node representation aggregation module, which generates a node representation of a current iteration process of the aggregate graph node according to the node representation of a previous iteration process of the aggregate graph node and the node representation of a previous iteration process of the neighbor graph nodes of the aggregate graph node, wherein the aggregate graph node includes the first graph node or the second graph node, and an initial node representation of the graph node is generated based on the node propagation representation of the graph node and the original features of the graph node, wherein the node representation propagation module and the node representation aggregation module execute a specified number of iterations.

[0016] Optionally, in an example of the above aspect, the node relationship representation generation unit concatenates the node representations of the first graph node and the second graph node to generate a node relationship representation between the first graph node and the second graph node.

[0017] Optionally, in an example of the above aspect, the node representation propagation module generates a node propagation representation of the current iteration process of the target graph node based on the node propagation representations received by each target graph node from each neighbor graph node, the edge relationship characteristics between each neighbor graph node and its own node propagation representation of the previous iteration process.

[0018] Optionally, in an example of the above aspect, the node representation aggregation module generates an initial node representation of the graph node by concatenating the node propagation representation of the graph node and the original features of the graph node.

[0019] Optionally, in an example of the above aspect, the device is implemented based on a graph neural network, and the node representation aggregation module uses an aggregation function to aggregate the node representations of the previous iteration process of the aggregate graph node and the node representations of the previous iteration process of the neighbor graph nodes of the aggregate graph node to generate a node representation of the current iteration process of the aggregate graph node.

[0020] According to another aspect of an embodiment of the present specification, there is provided a device for predicting a business relationship between a first graph node and a second graph node in graph data, wherein the graph data is generated based on business data, and the device comprises: a node representation determination unit, which starts node representation propagation and node representation aggregation from the first graph node and the second graph node, respectively, to determine the node representations of the second graph node and the first graph node; a node relationship representation generation unit, which generates a node relationship representation between the first graph node and the second graph node based on the node representations of the first graph node and the second graph node; and an inter-node business relationship prediction unit, which predicts the business relationship between the first graph node and the second graph node based on the node relationship representation between the first graph node and the second graph node, wherein the node representation determination unit comprises: a node representation propagation module, which propagates each of the source graph nodes in the current iteration process. The node propagation representation of the previous iteration process of the source graph node is propagated to each target graph node in the target graph node set of the source graph node, and the target graph node set is composed of the neighbor graph nodes of the source graph node, and the node propagation representation of the current iteration process of each target graph node is generated according to the node propagation representation received by each target graph node and its own node propagation representation of the previous iteration process; a node representation aggregation module generates a node representation of the current iteration process of the aggregate graph node according to the node representation of the previous iteration process of the aggregate graph node and the node representation of the previous iteration process of the neighbor graph nodes of the aggregate graph node, wherein the aggregate graph node includes the first graph node or the second graph node, and the initial node representation of the graph node is generated based on the node propagation representation of the graph node and the original feature of the graph node, wherein the node representation propagation module and the node representation aggregation module execute a specified number of iterations.

[0021] According to an embodiment of the present specification, there is provided a device for generating a node relationship representation between a first graph node and a second graph node in graph data, wherein the graph data is generated based on business data, and the device comprises: at least one processor, a memory coupled to the at least one processor, and a computer program stored in the memory, wherein the at least one processor executes the computer program to implement the graph node relationship representation generation method as described above.

[0022] According to an embodiment of the present specification, a computer-readable storage medium is provided, which stores executable instructions, and when the instructions are executed, the processor executes the graph node relationship representation generation method as described above.

[0023] According to an embodiment of the present specification, a computer program product is provided, including a computer program, wherein the computer program is executed by a processor to implement the method for generating a graph node relationship representation as described above.

[0024] According to an embodiment of the present specification, there is provided a device for predicting a relationship between a first graph node and a second graph node in graph data, wherein the graph data is generated based on business data, and the device comprises: at least one processor, a memory coupled to the at least one processor, and a computer program stored in the memory, wherein the at least one processor executes the computer program to implement the method for predicting business relationships between graph nodes as described above.

[0025] According to an embodiment of the present specification, a computer-readable storage medium is provided, which stores executable instructions, and when the instructions are executed, the processor executes the method for predicting business relationships between graph nodes as described above.

[0026] According to an embodiment of the present specification, a computer program product is provided, including a computer program, wherein the computer program is executed by a processor to implement the method for predicting business relationships between graph nodes as described above. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] A further understanding of the nature and advantages of the content of this specification may be achieved by referring to the following drawings. In the drawings, similar components or features may have the same reference numerals.

[0028] Figure 1 A schematic diagram of an example structure of graph data according to an embodiment of the present specification is shown.

[0029] Figure 2 An example flowchart of a method for generating a graph node relationship representation for generating a node relationship representation between a first graph node and a second graph node in graph data according to an embodiment of the present specification is shown.

[0030] Figure 3 An example schematic diagram of a node propagation process of graph data according to an embodiment of the present specification is shown.

[0031] Figure 4 An exemplary flowchart of a node representation propagation process according to an embodiment of the present specification is shown.

[0032] Figure 5An example schematic diagram of a node propagation representation generation process based on a graph neural network according to an embodiment of the present specification is shown.

[0033] Figure 6 An example schematic diagram of a node representation propagation process according to an embodiment of the present specification is shown.

[0034] Figure 7 An example schematic diagram of node representation aggregation processing according to an embodiment of the present specification is shown.

[0035] Figure 8 An example flow chart of a graph node business relationship prediction method for predicting a business relationship between a first graph node and a second graph node in graph data according to an embodiment of the present specification is shown.

[0036] Fig. 9 An example schematic diagram of a process for predicting business relationships between graph nodes according to an embodiment of the present specification is shown.

[0037] Fig.10 An example block diagram of a graph node relationship representation generating device for generating a node relationship representation between a first graph node and a second graph node in graph data according to an embodiment of the present specification is shown.

[0038] Fig.11 An example flow chart of a graph node business relationship prediction device for predicting a business relationship between a first graph node and a second graph node in graph data according to an embodiment of the present specification is shown.

[0039] Fig.12 An example schematic diagram of a graph node relationship representation generating device implemented based on a computer system for generating a node relationship representation between a first graph node and a second graph node in graph data according to an embodiment of the present specification is shown.

[0040] Fig.13 An example schematic diagram of a graph node business relationship prediction device for predicting a business relationship between a first graph node and a second graph node in graph data, implemented based on a computer system according to an embodiment of the present specification is shown. DETAILED DESCRIPTION

[0041] The subject matter described herein will now be discussed with reference to example embodiments. It should be understood that the discussion of these embodiments is only to enable those skilled in the art to better understand and implement the subject matter described herein, and is not a limitation of the scope of protection, applicability or examples set forth in the claims. The functions and arrangements of the elements discussed can be changed without departing from the scope of protection of the contents of this specification. Various examples can omit, replace or add various processes or components as needed. For example, the described method can be performed in an order different from the described order, and various steps can be added, omitted or combined. In addition, the features described relative to some examples can also be combined in other examples.

[0042] As used herein, the term "including" and its variations represent open terms, meaning "including but not limited to". The term "based on" means "based at least in part on". The terms "one embodiment" and "an embodiment" mean "at least one embodiment". The term "another embodiment" means "at least one other embodiment". The terms "first", "second", etc. may refer to different or the same objects. Other definitions may be included below, whether explicit or implicit. Unless the context clearly indicates otherwise, the definition of a term is consistent throughout the specification.

[0043] When business data is processed into graph data, it is necessary to predict the business relationship between two given graph nodes in the graph data during business processing. The business relationship prediction scheme may include, for example, a business relationship prediction scheme based on a graph structure, a business relationship prediction scheme based on a node-level graph neural network, and a business relationship prediction scheme based on an edge-level graph neural network.

[0044] In the business relationship prediction scheme based on graph structure, the characteristics of the two graph nodes to be predicted are directly calculated, such as the number of common neighbors, Katz Index, etc., so as to measure the similarity of the two graph nodes, and determine whether the two graph nodes have a business relationship based on the similarity. In this business relationship prediction scheme, certain a priori assumptions are required, that is, it is believed that there is a business relationship between two graph nodes that meet certain specific conditions. However, this business relationship prediction scheme cannot be adaptively adjusted according to different prediction tasks, and graph node attributes cannot be incorporated into node characterization.

[0045] In the business relationship prediction scheme based on node-level graph neural network, the graph neural network model is used to express the graph nodes of the two graph nodes to be predicted, and then the binary classification of the link is predicted based on the graph node expressions of the two graph nodes. However, this business relationship prediction scheme learns the graph node expressions of the two graph nodes independently, and can only model the neighbor distribution information of the two graph nodes, but cannot directly model the interaction structure (e.g., path) and interaction structure attributes between the two graph nodes.

[0046] In the business relationship prediction scheme based on edge-level graph neural network, first, the structural features are extracted for an edge in the graph data or all the domain graph nodes associated with an edge. For example, the connectivity features of an edge are extracted using random walks, and the positions of the domain graph nodes associated with an edge in the edge are represented using node representations. Next, the domain attribute information of the graph nodes is aggregated using a node-level graph neural network to obtain the final expression of the edge. Then, based on the final expression of the edge, it is predicted whether there is a business relationship between the two graph nodes connected by the edge. However, this business relationship prediction scheme cannot incorporate the graph node attribute information when extracting the structural features of the two graph nodes, and there are also prior assumptions in the selection of structural features.

[0047] In view of the above, an embodiment of the present specification provides a business relationship prediction scheme between graph nodes. In the graph node business relationship prediction scheme, when performing graph node relationship representation, node representation propagation and node representation aggregation are performed starting from one graph node to determine the node representation of the opposite graph node. Node representation propagation and node representation aggregation perform a specified number of iterations. By performing node representation propagation and node representation aggregation in an iterative process, the graph node attribute information and interaction structure information of each graph node can be integrated into the node representation of the opposite graph node. After obtaining the node representations of the two graph nodes as above, a graph node relationship representation between the two graph nodes is generated based on the node representations of the two graph nodes, and the business relationship between the nodes is predicted based on the node relationship representation between the two graph nodes. Using the above-mentioned prediction of the business relationship between nodes, since the graph node attribute information and interaction structure information of the two graph nodes are integrated into the node representation of the graph node, the obtained graph node relationship representation between the graph nodes can accurately reflect the node attribute information of the graph nodes and the interaction structure information between the graph nodes, thereby improving the accuracy of the prediction of the business relationship between the graph nodes.

[0048] The following will describe with reference to the accompanying drawings a graph node relationship representation generation method and device for generating a node relationship representation between a first graph node and a second graph node in graph data and a graph node business relationship prediction method and device for predicting a business relationship between a first graph node and a second graph node in graph data according to an embodiment of the present specification.

[0049] Figure 1 An example schematic diagram of a graph data structure according to an embodiment of the present specification is shown.

[0050] Figure 1 The graph data structure shown includes 11 graph nodes, namely, graph nodes u, v, and P0 to P8, and there are the following relationships between the 11 graph nodes: Figure 1 The neighbor relationship shown. In this specification, the term "neighbor relationship" refers to the existence of an edge connection between two graph nodes. For example, graph nodes P0, P1, P2, and P3 are 1-step neighbor graph nodes of graph node u, and graph node P4 is a 2-step neighbor graph node of graph node u. In addition, in this specification, unless otherwise specified, a neighbor graph node means a 1-step neighbor graph node.

[0051] In this specification, graph data can be generated based on various business data. Each graph node in the generated graph data has node attribute information. In addition, optionally, the edge between two graph nodes can also have edge attribute information (edge ​​features). The node attribute information and edge attribute information are related to the business scenario of the business data. Examples of business data may include, but are not limited to: social data, financial transaction data, commodity transaction data, and enterprise supply relationship data.

[0052] Social data includes business data obtained in social network scenarios. In the graph data generated based on social data, the graph nodes include social users (social accounts), such as "Zhang San", "Li Si", etc., and the node attribute information of the graph nodes may include, for example, age, education, address, occupation, etc. The edges between graph nodes may include, for example, the relationship between users, such as classmates / colleagues, etc. The predicted business relationship may include, for example, whether there is a friend relationship between social users.

[0053] Financial transaction data includes financial transaction data obtained in financial transaction scenarios. In graph data generated based on financial transaction data, graph nodes include financial entities, and node attribute information of graph nodes may include, for example, financial entity registration information, legal person information, account information, etc. Edges between graph nodes may include, for example, transfer relationships between financial entities, and predicted business relationships may include, for example, whether there is an associated entity relationship between financial entities.

[0054] Commodity transaction data includes commodity transaction data obtained in commodity transaction scenarios. In the graph data generated based on commodity transaction data, graph nodes may include users, commodities and merchants, and the node attribute information of the graph nodes may include, for example, user age, user education, user address, user occupation, commodity origin, commodity material, commodity model, merchant name, merchant registration place, etc. The edges between graph nodes may include, for example, commodity purchase relationships between users and commodities, and ownership relationships between commodities and merchants, etc. The predicted business relationships may include, for example, whether users have commodity purchase intentions.

[0055] The enterprise supply relationship data includes enterprise supply relationship data obtained in the enterprise supply scenario. In the graph data generated based on the enterprise supply relationship data, the graph nodes may include enterprises, and the node attributes of the graph nodes may include, for example, the enterprise name, the enterprise registration place, the goods produced by the enterprise, etc. The edges between the graph nodes may include, for example, the commodity supply behavior between the enterprises. The predicted business relationship may include, for example, whether there is a supply chain upstream and downstream relationship between two enterprises.

[0056] It should be noted that in this specification, the graph data structure may also be referred to as a graph network. In addition, examples of graph networks may include directed graph networks, undirected graph networks, and the like.

[0057] Figure 2 An example flow chart of a method 200 for generating a node relationship representation between a first graph node and a second graph node in graph data according to an embodiment of the present specification is shown. It should be noted that the first graph node and the second graph node may be any graph nodes in the graph data.

[0058] like Figure 2 As shown, at 210, node representation propagation and node representation aggregation are performed starting from the first graph node to determine the node representation of the second graph node. Here, the node representation propagation and node representation aggregation are performed for a specified number of iterations. In one example, the graph node relationship representation generation method can be implemented based on a graph neural network. In this case, the number of iterations of the iterative process performed by the node representation propagation and node representation aggregation can be the model depth of the graph neural network. The term "model depth" refers to the number of hops of graph nodes that each graph node can affect. That is, the number of hops of graph nodes to which the node attribute information of each graph node can be propagated. In one example, assuming that the model depth of the graph neural network is 2, the number of hops of graph nodes affected by each graph node is 2. In this case, for Figure 1In the graph network shown in , the first graph node is graph node u, and the second graph node is graph node v. The graph nodes to which the node attribute information of graph node u can be propagated include graph nodes P0, P1, P2, P3, P4, P5 and P6, wherein graph nodes P0, P1, P2 and P3 are one-hop target graph nodes (1-step neighbor graph nodes) of graph node u, and graph nodes P4, P5 and P6 are two-hop target graph nodes (2-step neighbor graph nodes) of graph node u. In addition, the node attribute information of graph node u cannot be propagated to graph nodes beyond 2 hops, for example, graph nodes v, P7 and P8. In this specification, examples of graph neural networks may include, for example, but are not limited to, graph convolutional networks (GCNs), graph attention networks (GATs), etc.

[0059] Figure 3 An example schematic diagram of a node propagation process of graph data according to an embodiment of the present specification is shown.

[0060] like Figure 3 As shown, in the first iteration of the node representation propagation starting from graph node u, the source graph node is graph node u, and the target graph nodes are graph nodes P0, P1, P2, and P3. During the first iteration, the node propagation features of graph node u are propagated to graph nodes P0, P1, P2, and P3, respectively. During the second iteration, the source graph nodes are graph nodes P0, P1, P2, and P3, and the target graph nodes are graph nodes P4, P5, and P6. During the second iteration, the node propagation features of source graph node P2 are propagated to graph nodes P4 and P5, respectively, and the node propagation features of source graph node P3 are propagated to graph nodes P5 and P6, respectively.

[0061] Figure 4 An example flow chart of a node representation propagation process 400 according to an embodiment of the present specification is shown.

[0062] like Figure 4 As shown, at 410, the node propagation representation of the previous iteration process of each source graph node in the source graph node set of the current iteration process is propagated to each target graph node in the target graph node set of the source graph node, and the target graph node set is composed of the 1-step neighbor graph nodes of the source graph node.

[0063] At 420 , a node propagation representation of a current iteration process of each target graph node is generated according to the node propagation representation received by each target graph node from the neighbor graph node and its own node propagation representation of a previous iteration process.

[0064] Figure 5 FIG. 5 shows an example schematic diagram of a node propagation representation generation process 500 based on a graph neural network according to an embodiment of the present specification. Figure 5In the example, the graph neural network includes a graph neural network with an attention mechanism and an LSTM aggregator.

[0065] like Figure 5 As shown, at 510, the node propagation representations received by each target graph node from the neighbor graph nodes are aggregated through the Attention operation (ATT_OP) to obtain the neighbor node propagation representations of each target graph node.

[0066] In one example, the processing of ATT_OP can be expressed using the following expression:

[0067]

[0068] in, is a given parameter in the graph neural network model, and is the weight coefficient matrix in the graph neural network model, is the attention value propagated from the source graph node q to the target graph node p, is the node propagation representation of the source graph node q in the k-1th round of iteration, X p is the original feature of the target graph node p, and It represents the propagation representation of the neighbor nodes of the target graph node p after k rounds of iterations starting from the graph node u.

[0069] In addition, in one example, an edge relationship feature e is also included between two graph nodes q and p q,p In the case of The calculation formula can be modified as follows:

[0070]

[0071] At 520, the neighbor node propagation representations of each target graph node and the node propagation representation of the previous iteration process of the target graph node are subjected to LSTM aggregation (LSTM_OP) by an LSTM aggregator to generate the node propagation representation of the current iteration process of each target graph node.

[0072] In an example, the processing of LSTM_OP can be expressed by the following expression:

[0073]

[0074]

[0075] in, is the input gate, For the Gate of Forgetfulness, is the output gate, is the cell state, and The node propagation representation of the target graph node p after k-1 and k rounds of iterations starts from the graph node u.

[0076] Below Figure 1 The graph network example shown in FIG. 1 is used as an example to describe the node representation propagation process. Figure 6 FIG. 2 shows an example schematic diagram of node representation propagation processing according to an embodiment of the present specification. Figure 6 In the example shown, the model depth of the graph neural network is k=2.

[0077] In the first iteration process (k=1), the source graph node is graph node u, and the target graph nodes are graph nodes P0, P1, P2, and P3. Accordingly, the node propagation representation of graph node u is propagated to graph nodes P0, P1, P2, and P3 respectively. Here, the node propagation feature of graph node u is equivalent to the original feature information X u In one example, the original feature information X of each graph node is u It can be obtained by embedding the node attribute information of the graph node. Through the above processing, the original feature information of each graph node can be represented as a vector of a specified dimension, and the value of each dimension is a real value between [0,1]. For example, when the specified dimension is 10, the original feature information X of the graph node u is u For example, it can be represented as a 10-dimensional vector [0.12, 0.43, 0.35, 0.33, 0.28, 0.59, 0.77, 0.94, 0.66, 0.84].

[0078] At each graph node P0, P1, P2, and P3, the node propagation representations received by each graph node P0, P1, P2, and P3 from the neighboring graph node u are aggregated by ATT_OP as above to obtain the neighboring node propagation representations of each graph node P0, P1, P2, and P3. and

[0079] Then, at each graph node P0, P1, P2, and P3, the neighbor nodes of each graph node P0, P1, P2, and P3 are propagated with the representation through the LSTM aggregator. and And the node propagation representation of the previous iteration process of the graph node and Perform LSTM_OP to generate node propagation representations of the first iteration of each graph node P0, P1, P2, and P3 and In the first iteration, and The value of is 0.

[0080] In the second iteration (k=2), the source graph nodes are graph nodes P0, P1, P2, and P3, and the target graph nodes are graph nodes P4, P5, and P6. Accordingly, the node propagation representation of graph node P2 is Propagate to graph nodes P4 and P5, and propagate the node representation of graph node P3 Propagate to graph nodes P5 and P6. Since graph nodes P0 and P1 have no neighboring graph nodes, graph nodes P0 and P1 do not undergo further node propagation processing.

[0081] At each graph node P4 and P5, the node propagation representations received by each graph node P4 and P5 from the neighboring graph nodes are aggregated by ATT_OP as above to obtain the neighboring node propagation representations of each graph node P4 and P5. and Specifically, at graph node P4, the node propagation representation of graph node P2 is aggregated by ATT_OP as above Get the neighbor node propagation representation of graph node P4 At graph node P5, the node propagation representation of graph nodes P2 and P3 is aggregated through ATT_OP as above and Get the neighbor node propagation representation of graph node P5 At graph node P6, the node propagation representation of graph node P3 is aggregated through ATT_OP as above Get the neighbor node propagation representation of graph node P6

[0082] Then, at each graph node P4, P5, and P6, the neighbor nodes of each graph node P4, P5, and P6 are propagated with the representation through the LSTM aggregator. and And the node propagation representation of the previous iteration process of the graph node and Perform LSTM_OP to generate node propagation representations of the second iteration process for each graph node P4, P5, and P6 and exist Figure 6 In the example, since the node representation propagation processing of graph nodes P4, P5 and P6 is not involved in the first iteration, and The value of is 0.

[0083] After obtaining the node propagation representation of each graph node through the above iterative processing, node representation aggregation is performed. Similarly, node representation aggregation needs to perform a specified number of iterations.

[0084] In each iteration of node representation aggregation, a node representation of the current iteration of the second graph node is generated based on the node representation of the previous iteration of the second graph node and the node representation of the previous iteration of the neighboring graph nodes of the second graph node. Here, the initial node representation of the graph node involved in the node representation aggregation is generated based on the node propagation representation of the graph node and the original feature of the graph node.

[0085] Figure 7 An example schematic diagram of a node representation aggregation process according to an embodiment of the present specification is shown. Figure 7 The example of FIG. 1 shows node representation aggregation for graph node v.

[0086] For all the Graph nodes (for example, when H = 2 Figure 1 There are two different types of graph nodes, one of which is (intersection), and another type of graph node is (difference). Here, represents the H-step neighbors centered at the graph node v, and represents the H-step neighbors centered on the graph node u. When performing node representation aggregation, first determine the initial node representation of each graph node p In one example, the initial node representation of a graph node is generated by concatenating the node propagation representation of the graph node and the original features of the graph node. For example, the initial node representation of graph node p is It can be expressed by the following formula:

[0087]

[0088] After obtaining the initial node representation of each graph node as above, the node representation of graph node p in each iteration process is determined according to the iteration process. In one example, assuming that in the kth iteration process, the node representation of graph node p The aggregation function AGG(.) in the graph neural network can be used to determine as follows:

[0089]

[0090] Among them, AGG(.) is the aggregation function in the graph neural network, is the node representation of the graph node p in the k-th iteration process, is the node representation of the graph node p in the k-1th iteration process, represents the edge between H-step neighbor nodes centered on graph node v, (i,p) represents the edge from graph node i to graph node p, is the node representation of graph node i during the k-1th iteration.

[0091] Iterate the node representation aggregation in the above way, and finally obtain the H-order node representation of the graph node v (H iterations) As the final node representation of graph node v.

[0092] Back to Figure 2 At 220, node representation propagation and node representation aggregation are performed starting from the second graph node to determine the node representation of the first graph node. The processing of step 220 is similar to the processing of step 210, except that node representation propagation and node representation aggregation are started from graph node v and the node representation of graph node u is finally calculated, which will not be described here. Thus, the H-order node representation of graph node u is obtained (H iterations) As the final node representation of the graph node u. It should be noted that in this specification, the first graph node or the second graph node may be referred to as an aggregated graph node.

[0093] At 230, a node relationship representation between the first graph node and the second graph node is generated based on the node representations of the first graph node and the second graph node. In one example, the node relationship representation between the first graph node and the second graph node can be generated by concatenating the node representations of the first graph node and the second graph node. For example, for graph nodes u and v, the node representation of graph node u can be concatenated to generate the node relationship representation between the first graph node and the second graph node. and the node representation of the graph node v Get the node relationship representation of graph node u and graph node v

[0094] Figure 8 An example flow chart of a graph node business relationship prediction method 800 for predicting a business relationship between a first graph node and a second graph node in graph data according to an embodiment of the present specification is shown. Figure 8 In the example, operations from 810 to 830 are exactly the same as operations from 210 to 230 above and will not be described again here.

[0095] After obtaining the node relationship representation between the first graph node and the second graph node as above, graph node business relationship prediction is performed based on the node relationship representation between the first graph node and the second graph node at 840. The graph node business relationship prediction process based on the node relationship representation can be implemented using any prediction method applicable in the art.

[0096] For example, the graph node business relationship prediction process based on the node relationship representation can be implemented using a machine learning model such as a graph neural network. In one example, when training the machine learning model, the following function can be used as the loss function of the model training.

[0097]

[0098] in, is the model prediction result, y = MLP (s u,v ), MLP(.) is a multi-layer perceptron, y∈{0,1}, when<u,v> When there is a business relationship between them, y=1, otherwise y=0. is the training sample data set, Represents the size of the training sample data set, that is, the number of training sample data.

[0099] Fig. 9 An example schematic diagram of a node business relationship prediction process based on a graph neural network according to an embodiment of the present specification is shown. Fig. 9 The example shows the node business relationship prediction process between graph node u and graph node v.

[0100] like Fig. 9 As shown, first, node representation propagation and node representation aggregation are performed for a specified number of iterations starting from graph node u to determine the node representation of graph node v. And starting from the graph node v, node representation propagation and node representation aggregation are performed for a specified number of iterations to determine the node representation of the graph node u Then, the node representation of the graph node v is and the node representation of graph node u Splice and get the node relationship representation s of graph node u and graph node v u,v Here, the node representations of graph nodes u and v are And the node relationship representation s between graph node u and graph node v u,v It can be represented as a representation vector. The representation vector can include multiple dimensions, and the dimension value of each dimension can be represented by a positive real value, for example, a value between [0, 1]. Node relationship representation s u,v The fully connected layer (FC layer) of the graph neural network is provided to perform business relationship prediction. Subsequently, whether there is a business relationship between the graph nodes is predicted based on the business relationship prediction value. For example, the obtained business relationship prediction value can be a probability value. When the obtained business relationship prediction value is not less than a predetermined threshold, it is considered that there is a business relationship between the graph nodes. When the obtained business relationship prediction value is less than a predetermined threshold, it is considered that there is no business relationship between the graph nodes.

[0101] Using the above-mentioned prediction of business relationships between nodes, node representation propagation and node representation aggregation are performed in the process of determining the node representation of the graph node. The graph node representation obtained in this way incorporates the graph node attribute information and interaction structure information of the two graph nodes, so that the obtained graph node relationship representation between the graph nodes can accurately reflect the node attribute information of the graph nodes and the interaction structure information between the graph nodes, thereby improving the accuracy of the prediction of business relationships between the graph nodes.

[0102] In addition, by utilizing the above-mentioned prediction of business relationships between nodes, by also propagating the edge relationship feature data between two graph nodes when propagating node representations, the obtained graph node relationship representations between graph nodes can also reflect the edge feature information between graph nodes, thereby further improving the accuracy of business relationship predictions between graph nodes.

[0103] Fig.10 FIG. 1 is a block diagram showing an example of a graph node relationship representation generating device 1000 for generating a node relationship representation between a first graph node and a second graph node in graph data according to an embodiment of the present specification. Fig.10 As shown, the graph node relationship representation generating device 1000 includes a node representation determining unit 1010 and a node relationship representation generating unit 1020 .

[0104] The node representation determination unit 1010 is configured to perform node representation propagation and node representation aggregation starting from the first graph node and the second graph node, respectively, to determine the node representations of the second graph node and the first graph node. The operation of the node representation determination unit 1010 may refer to the above reference. Figure 2 The operations described in 210 and 220 .

[0105] The node relationship representation generating unit 1020 is configured to generate a node relationship representation between the first graph node and the second graph node based on the node representations of the first graph node and the second graph node. The operation of the node relationship representation generating unit 1020 can refer to the above reference. Figure 2 The operation described in 230 .

[0106] The node representation determination unit 1010 includes a node representation propagation module 1011 and a node representation aggregation module 1013. The node representation propagation module 1011 and the node representation aggregation module 1013 are configured to perform a specified number of iterations to determine the node representations of the first graph node and the second graph node respectively.

[0107] Specifically, in each iteration process, the node representation propagation module 1011 propagates the node propagation representation of the previous iteration process of each source graph node in the source graph node set of the current iteration process to each target graph node in the target graph node set of the source graph node, wherein the target graph node set is composed of the neighboring graph nodes of the source graph node, and generates the node propagation representation of the current iteration process of each target graph node based on the node propagation representation received by each target graph node and its own node propagation representation of the previous iteration process.

[0108] In each iteration process, the node representation aggregation module 1013 is configured to generate a node representation of the current iteration process of the aggregate graph node based on the node representation of the previous iteration process of the aggregate graph node and the node representation of the previous iteration process of the neighbor graph node of the aggregate graph node, wherein the aggregate graph node includes the first graph node or the second graph node, and the initial node representation of the graph node is generated based on the node propagation representation of the graph node and the original feature of the graph node. It should be noted that the iteration process of the node representation aggregation module 1013 is executed after the iteration process of the node representation propagation module 1011 is completed.

[0109] In addition, in some embodiments, the node relationship representation generation unit 1020 concatenates the node representations of the first graph node and the second graph node to generate a node relationship representation between the first graph node and the second graph node.

[0110] In addition, in some embodiments, when there are edge relationship features between graph nodes, the node representation propagation module 1011 is configured to generate a node propagation representation of the current iteration process of the target graph node based on the node propagation representations received by each target graph node from each neighbor graph node, the edge relationship features between each neighbor graph node, and the node propagation representation of the previous iteration process.

[0111] In addition, in some embodiments, the node representation aggregation module 1013 generates an initial node representation of the graph node by concatenating the node propagation representation of the graph node and the original features of the graph node.

[0112] In addition, in some embodiments, when the graph node relationship representation generating device 1000 is implemented based on a graph neural network, the node representation aggregation module 1013 is configured to use an aggregation function to aggregate the node representations of the previous iteration process of the aggregate graph node and the node representations of the previous iteration process of the neighbor graph nodes of the aggregate graph node to generate a node representation of the current iteration process of the aggregate graph node.

[0113] Fig.11 FIG. 1 shows an example flow chart of a graph node business relationship prediction device 1100 for predicting a business relationship between a first graph node and a second graph node in graph data according to an embodiment of the present specification. Fig.11 As shown, the graph node business relationship prediction device 1100 includes a node representation determination unit 1110, a node relationship representation generation unit 1120, and an inter-node business relationship prediction unit 1130. The operation and structure of the node representation determination unit 1110 and the node relationship representation generation unit 1120 are the same as the operation and structure of the node representation determination unit 1010 and the node relationship representation generation unit 1020 described above, and will not be repeated here.

[0114] The node-to-node business relationship prediction unit 1130 is configured to predict the business relationship between the first graph node and the second graph node according to the node relationship representation between the first graph node and the second graph node. The operation of the node-to-node business relationship prediction unit 1130 can refer to the above reference. Figure 8 The operation of 840 is described.

[0115] As above Figures 1 to 11 , a graph node relationship representation generation method, a graph node relationship representation generation device, a graph node business relationship prediction method, and a graph node business relationship prediction device according to an embodiment of this specification are described. The above-mentioned graph node relationship representation generation device and graph node business relationship prediction device can be implemented by hardware, or by software, or a combination of hardware and software.

[0116] Fig.12 A schematic diagram of a graph node relationship representation generating device 1200 for generating a node relationship representation between a first graph node and a second graph node in graph data, implemented based on a computer system according to an embodiment of the present specification is shown. Fig.12 As shown, the graph node relationship representation generating device 1200 may include at least one processor 1210, a memory (e.g., a non-volatile memory) 1220, a memory 1230, and a communication interface 1240, and the at least one processor 1210, the memory 1220, the memory 1230, and the communication interface 1240 are connected together via a bus 1260. At least one processor 1210 executes at least one computer-readable instruction stored or encoded in the memory (i.e., the above-mentioned element implemented in the form of software).

[0117] In one embodiment, computer executable instructions are stored in a memory, which, when executed, cause at least one processor 1210 to: perform node representation propagation and node representation aggregation starting from a first graph node and a second graph node, respectively, to determine the node representations of the second graph node and the first graph node; and generate a node relationship representation between the first graph node and the second graph node based on the node representations of the first graph node and the second graph node, wherein the node representation propagation and node representation aggregation are performed for a specified number of iterations, and during the node representation propagation, the node propagation representation of the previous iteration process of each source graph node in the source graph node set of the current iteration process is propagated to the target graph node set of the source graph node for each target graph node, the target graph node set consists of the neighbor graph nodes of the source graph node; generating the node propagation representation of the current iteration process of each target graph node according to the node propagation representation received by each target graph node and its own node propagation representation of the previous iteration process, and when the node representations are aggregated, generating the node representation of the current iteration process of the aggregate graph node according to the node representation of the previous iteration process of the aggregate graph node and the node representation of the neighbor graph nodes of the aggregate graph node in the previous iteration process, wherein the aggregate graph node includes the first graph node or the second graph node, and the initial node representation of the graph node is generated based on the node propagation representation of the graph node and the original feature of the graph node.

[0118] It should be understood that the computer executable instructions stored in the memory, when executed, cause at least one processor 1210 to perform the above combined operations in various embodiments of the present specification. Figure 1-Figure 7 as well as Fig.10 Describes the various operations and functions.

[0119] Fig.13 FIG. 1 is a schematic diagram of a graph node business relationship prediction device 1300 implemented based on a computer system for predicting a business relationship between a first graph node and a second graph node in graph data according to an embodiment of the present specification. Fig.13 As shown, the graph node business relationship prediction device 1300 may include at least one processor 1310, a memory (e.g., a non-volatile memory) 1320, a memory 1330, and a communication interface 1340, and the at least one processor 1310, the memory 1320, the memory 1330, and the communication interface 1340 are connected together via a bus 1360. At least one processor 1310 executes at least one computer-readable instruction stored or encoded in the memory (i.e., the above-mentioned element implemented in the form of software).

[0120] In one embodiment, computer executable instructions are stored in a memory, which, when executed, cause at least one processor 1310 to: perform node representation propagation and node representation aggregation starting from a first graph node and a second graph node, respectively, to determine node representations of the second graph node and the first graph node; generate a node relationship representation between the first graph node and the second graph node based on the node representations of the first graph node and the second graph node; and predict a business relationship between the first graph node and the second graph node based on the node relationship representation between the first graph node and the second graph node, wherein the node representation propagation and node representation aggregation are performed for a specified number of iterations, and during the node representation propagation, the node representation of each source graph node in the source graph node set of the current iteration is transmitted to the source graph node set of the current iteration. The point propagation representation is propagated to each target graph node in the target graph node set of the source graph node, wherein the target graph node set is composed of the neighbor graph nodes of the source graph node; the node propagation representation of the current iteration process of each target graph node is generated according to the node propagation representation received by each target graph node and its own node propagation representation of the previous iteration process, and when the node representations are aggregated, the node representation of the current iteration process of the aggregate graph node is generated according to the node representation of the previous iteration process of the aggregate graph node and the node representation of the previous iteration process of the neighbor graph nodes of the aggregate graph node, wherein the aggregate graph node includes the first graph node or the second graph node, and the initial node representation of the graph node is generated based on the node propagation representation of the graph node and the original feature of the graph node.

[0121] It should be understood that the computer executable instructions stored in the memory, when executed, cause at least one processor 1310 to perform the above combined Figure 8-Figure 9 as well as Fig.11 Describes the various operations and functions.

[0122] According to one embodiment, a program product such as a machine-readable medium (e.g., a non-transitory machine-readable medium) is provided. The machine-readable medium may have instructions (i.e., the above-mentioned elements implemented in the form of software), which, when executed by a machine, causes the machine to perform the above-mentioned combination of various embodiments of this specification. Figure 1-Figure 11 Specifically, a system or device equipped with a readable storage medium may be provided, on which a software program code implementing the functions of any of the above-mentioned embodiments is stored, and a computer or processor of the system or device reads and executes the instructions stored in the readable storage medium.

[0123] In this case, the program code itself read from the machine-readable medium can realize the function of any one of the above-mentioned embodiments, and thus the machine-readable code and the machine-readable storage medium storing the machine-readable code constitute a part of the present invention.

[0124] Examples of readable storage media include floppy disks, hard disks, magneto-optical disks, optical disks (such as CD-ROM, CD-R, CD-RW, DVD-ROM, DVD-RAM, DVD-RW, DVD-RW), magnetic tapes, non-volatile memory cards, and ROMs. Alternatively, the program code may be downloaded from a server computer or a cloud via a communication network.

[0125] According to one embodiment, a computer program product is provided. The computer program product includes a computer program. When the computer program is executed by a processor, the processor performs the above combination of the various embodiments of this specification. Figure 1-Figure 11 Describes the various operations and functions.

[0126] Those skilled in the art should understand that the various embodiments disclosed above can be modified and altered in various ways without departing from the essence of the invention. Therefore, the protection scope of the present invention should be defined by the appended claims.

[0127] It should be noted that not all steps and units in the above-mentioned processes and system structure diagrams are necessary, and some steps or units can be ignored according to actual needs. The execution order of each step is not fixed and can be determined as needed. The device structure described in the above-mentioned embodiments can be a physical structure or a logical structure, that is, some units may be implemented by the same physical entity, or some units may be implemented by multiple physical entities, or some components in multiple independent devices may be implemented together.

[0128] In the above embodiments, the hardware unit or module can be realized by mechanical or electrical means. For example, a hardware unit, module or processor can include permanent dedicated circuit or logic (such as special processor, FPGA or ASIC) to complete the corresponding operation. The hardware unit or processor can also include programmable logic or circuit (such as general-purpose processor or other programmable processor), which can be temporarily set by software to complete the corresponding operation. Specific implementation (mechanical method or dedicated permanent circuit or temporary circuit) can be determined based on cost and time consideration.

[0129] The specific embodiments described above in conjunction with the accompanying drawings describe exemplary embodiments, but do not represent all embodiments that can be implemented or fall within the scope of protection of the claims. The term "exemplary" used throughout this specification means "used as an example, instance or illustration" and does not mean "preferred" or "having advantages" over other embodiments. For the purpose of providing an understanding of the described technology, the specific embodiments include specific details. However, these technologies can be implemented without these specific details. In some instances, in order to avoid making the concepts of the described embodiments difficult to understand, well-known structures and devices are shown in block diagram form.

[0130] The above description of the present disclosure is provided to enable any person of ordinary skill in the art to implement or use the present disclosure. Various modifications to the present disclosure will be apparent to those of ordinary skill in the art, and the general principles defined herein may be applied to other variations without departing from the scope of protection of the present disclosure. Therefore, the present disclosure is not limited to the examples and designs described herein, but is consistent with the widest range of principles and novel features disclosed herein.

Claims

1. A method for generating a node relationship representation between a first graph node and a second graph node in graph data, wherein the graph data is generated based on business data, the method comprising: Perform node representation propagation and node representation aggregation starting from the first graph node and the second graph node respectively, and determine the node representations of the second graph node and the first graph node; as well as generating a node relationship representation between the first graph node and the second graph node based on the node representations of the first graph node and the second graph node, The node representation propagation and the node representation aggregation are iterated for a specified number of times. When the node representation propagates, Propagate the node propagation representation of the previous iteration process of each source graph node in the source graph node set of the current iteration process to each target graph node in the target graph node set of the source graph node, wherein the target graph node set is composed of neighbor graph nodes of the source graph node; According to the node propagation representation received by each target graph node and its own node propagation representation of the previous iteration process, the node propagation representation of the current iteration process of each target graph node is generated. When the node representations are aggregated, a node representation of the current iteration process of the aggregate graph node is generated based on the node representation of the previous iteration process of the aggregate graph node and the node representation of the previous iteration process of the neighbor graph node of the aggregate graph node, wherein the aggregate graph node includes the first graph node or the second graph node, and the initial node representation of the graph node is generated based on the node propagation representation of the graph node and the original features of the graph node.

2. The method of claim 1, wherein: Generating a node relationship representation between the first graph node and the second graph node based on the node representations of the first graph node and the second graph node includes: The node representations of the first graph node and the second graph node are concatenated to generate a node relationship representation between the first graph node and the second graph node.

3. The method of claim 1, wherein: According to the node propagation representation received by each target graph node and its own node propagation representation of the previous iteration process, the node propagation representation of the current iteration process of the target graph node is generated, including: The node propagation representation of the current iteration process of the target graph node is generated according to the node propagation representation received by each target graph node from each neighbor graph node, the edge relationship characteristics between each neighbor graph node and its own node propagation representation of the previous iteration process.

4. The method of claim 1, wherein: The initial node representation of the graph node is generated based on the node propagation representation of the graph node and the original features of the graph node, including: The initial node representation of a graph node is generated by concatenating the node propagation representation of the graph node and the original features of the graph node.

5. The method of claim 1, wherein: The method is implemented based on graph neural network.

6. The method of claim 5, wherein: Generating the node representation of the current iteration process of the aggregate graph node according to the node representation of the previous iteration process of the aggregate graph node and the node representation of the previous iteration process of the neighbor graph node of the aggregate graph node includes: The node representation of the previous iteration process of the aggregate graph node and the node representation of the previous iteration process of the neighbor graph nodes of the aggregate graph node are aggregated using an aggregation function to generate a node representation of the current iteration process of the aggregate graph node.

7. The method of claim 6, wherein: The graph neural network includes a graph neural network with an Attention mechanism and an LSTM aggregator.

8. The method of claim 7, wherein: According to the node propagation representation received by each target graph node and its own node propagation representation of the previous iteration process, the node propagation representation of the current iteration process of each target graph node is generated, including: Aggregate the node propagation representations received by each target graph node through the Attention operation to obtain the neighbor node propagation representations of each target graph node; and The neighbor node propagation representation of each target graph node and the node propagation representation of the previous iteration process of the target graph node are aggregated by the LSTM aggregator to generate the node propagation representation of the current iteration process of each target graph node.

9. The method of claim 1, wherein: The business data includes one of the following business data: Social data; Financial transaction data; Commodity trading data; and Enterprise supply relationship data.

10. A method for predicting a business relationship between a first graph node and a second graph node in graph data, wherein the graph data is generated based on business data, the method comprising: Perform node representation propagation and node representation aggregation starting from the first graph node and the second graph node respectively, and determine the node representations of the second graph node and the first graph node; generating a node relationship representation between the first graph node and the second graph node based on the node representations of the first graph node and the second graph node; as well as predicting a business relationship between the first graph node and the second graph node according to a node relationship representation between the first graph node and the second graph node, The node representation propagation and the node representation aggregation are iterated for a specified number of times. When the node representation propagates, Propagate the node propagation representation of the previous iteration process of each source graph node in the source graph node set of the current iteration process to each target graph node in the target graph node set of the source graph node, wherein the target graph node set is composed of neighbor graph nodes of the source graph node; According to the node propagation representation received by each target graph node and its own node propagation representation of the previous iteration process, the node propagation representation of the current iteration process of each target graph node is generated. When the node representations are aggregated, a node representation of the current iteration process of the aggregate graph node is generated based on the node representation of the previous iteration process of the aggregate graph node and the node representation of the previous iteration process of the neighbor graph node of the aggregate graph node, wherein the aggregate graph node includes the first graph node or the second graph node, and the initial node representation of the graph node is generated based on the node propagation representation of the graph node and the original features of the graph node.

11. A device for generating a node relationship representation between a first graph node and a second graph node in graph data, wherein the graph data is generated based on business data, the device comprising: A node representation determination unit, performing node representation propagation and node representation aggregation from the first graph node and the second graph node, respectively, to determine the node representations of the second graph node and the first graph node; as well as a node relationship representation generating unit, which generates a node relationship representation between the first graph node and the second graph node based on the node representations of the first graph node and the second graph node, Wherein, the node characterization determination unit includes: A node representation propagation module propagates the node propagation representation of the previous iteration process of each source graph node in the source graph node set of the current iteration process to each target graph node in the target graph node set of the source graph node, wherein the target graph node set is composed of neighbor graph nodes of the source graph node, and generates the node propagation representation of the current iteration process of each target graph node according to the node propagation representation received by each target graph node and its own node propagation representation of the previous iteration process; A node representation aggregation module generates a node representation of a current iteration process of the aggregate graph node according to the node representation of the previous iteration process of the aggregate graph node and the node representation of the previous iteration process of the neighbor graph node of the aggregate graph node, wherein the aggregate graph node includes the first graph node or the second graph node, and the initial node representation of the graph node is generated based on the node propagation representation of the graph node and the original feature of the graph node. The node representation propagation module and the node representation aggregation module perform a specified number of iterations.

12. The device of claim 11, wherein: The node relationship representation generating unit concatenates the node representations of the first graph node and the second graph node to generate a node relationship representation between the first graph node and the second graph node.

13. The device of claim 11, wherein: The node representation propagation module generates a node propagation representation of the current iteration process of the target graph node based on the node propagation representations received by each target graph node from each neighbor graph node, the edge relationship characteristics between each neighbor graph node and its own node propagation representation of the previous iteration process.

14. The device of claim 11, wherein: The node representation aggregation module generates an initial node representation of the graph node by concatenating the node propagation representation of the graph node and the original features of the graph node.

15. The device of claim 11, wherein: The device is implemented based on a graph neural network. The node representation aggregation module uses an aggregation function to aggregate the node representations of the previous iteration process of the aggregate graph node and the node representations of the previous iteration process of the neighbor graph nodes of the aggregate graph node to generate a node representation of the current iteration process of the aggregate graph node.

16. A device for predicting a business relationship between a first graph node and a second graph node in graph data, wherein the graph data is generated based on business data, the device comprising: A node representation determination unit, performing node representation propagation and node representation aggregation from the first graph node and the second graph node, respectively, to determine the node representations of the second graph node and the first graph node; a node relationship representation generating unit, which generates a node relationship representation between the first graph node and the second graph node based on the node representations of the first graph node and the second graph node; as well as an inter-node business relationship prediction unit, which predicts the business relationship between the first graph node and the second graph node according to the node relationship representation between the first graph node and the second graph node, Wherein, the node characterization determination unit includes: A node representation propagation module propagates the node propagation representation of the previous iteration process of each source graph node in the source graph node set of the current iteration process to each target graph node in the target graph node set of the source graph node, wherein the target graph node set is composed of neighbor graph nodes of the source graph node, and generates the node propagation representation of the current iteration process of each target graph node according to the node propagation representation received by each target graph node and its own node propagation representation of the previous iteration process; A node representation aggregation module generates a node representation of a current iteration process of the aggregate graph node according to the node representation of the previous iteration process of the aggregate graph node and the node representation of the previous iteration process of the neighbor graph node of the aggregate graph node, wherein the aggregate graph node includes the first graph node or the second graph node, and the initial node representation of the graph node is generated based on the node propagation representation of the graph node and the original feature of the graph node. The node representation propagation module and the node representation aggregation module perform a specified number of iterations.

17. A device for generating a node relationship representation between a first graph node and a second graph node in graph data, wherein the graph data is generated based on business data, the device comprising: at least one processor, a memory coupled to the at least one processor, and A computer program stored in the memory, wherein the at least one processor executes the computer program to implement the method according to any one of claims 1 to 9.

18. A device for predicting a business relationship between a first graph node and a second graph node in graph data, wherein the graph data is generated based on business data, the device comprising: at least one processor, a memory coupled to the at least one processor, and A computer program stored in the memory, the at least one processor executing the computer program to implement the method of claim 10.

19. A computer-readable storage medium storing executable instructions, which, when executed, cause a processor to perform the method according to any one of claims 1 to 10.

20. A computer program product, comprising a computer program, wherein the computer program is executed by a processor to implement the method according to any one of claims 1 to 10.

Images