Data graph generation method and device, program product, storage medium and electronic equipment

Abstract

The embodiment of the invention provides a data graph generation method and device, a program product, a storage medium and electronic equipment, and the method comprises the steps: obtaining node information of N nodes in a target region; initial vector conversion operation is executed on the N pieces of node information, N initial feature vectors are obtained, and the initial feature vectors are used for representing the node information in a vector mode; a pre-trained first target network is used for executing graph structure conversion operation on the N initial feature vectors to obtain a structure relation graph between the N nodes, the structure relation graph comprises the N nodes, the connection relation between the N nodes and the node types of the N nodes, and the N is a positive integer; according to the method and the device, the problem that the structural relationship among a plurality of nodes cannot be accurately determined in the prior art is solved, and the technical effects that the structural relationship among the nodes can be accurately determined and the structural relationship among the nodes can be displayed in the structural relationship graph are achieved.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of data processing, and specifically, to a data graph generation method and device, a program product, a storage medium, and an electronic device. Background Art

[0002] In the modern information society, communication, navigation and remote sensing technologies play an important role in many fields such as civil aviation, navigation, urban planning, and environmental monitoring. However, communication data, navigation data, and remote sensing data each have different characteristics. For example, communication data includes text and signaling information, navigation data focuses on location and path information, and remote sensing data focuses on geographic images and environmental monitoring information.

[0003] Related technologies mainly use adaptive carrier modulation and coding technology to modulate short messages, remote sensing images, geographic location information and other different data at the data transmission layer, or process geographic coordinates and geographic information to build a graph structure. However, since different types of data have different data formats and semantics, and the data sources are diverse, the structure is complex, and exists in different forms, the generated graph structure cannot show the relationship between nodes.

[0004] To address the above-mentioned problems, no effective solution has been proposed yet. Summary of the invention

[0005] The embodiments of the present application provide a data graph generation method and device, a program product, a storage medium and an electronic device to at least solve the problem in the related art that the structural relationship between multiple nodes cannot be accurately determined.

[0006] According to one embodiment of the present application, a data graph generation method is provided, comprising: obtaining node information of N nodes in a target area, wherein the above N is a natural number greater than 1; performing an initial vector conversion operation on the N above node information to obtain N initial feature vectors, wherein the above initial feature vectors are used to represent the above node information in the form of a vector; using a pre-trained first target network to perform a graph structure conversion operation on the N above initial feature vectors to obtain a structural relationship graph between the N above nodes, wherein the above structural relationship graph includes: the N above nodes, the connection relationship between the N above nodes, and the node types of the N above nodes.

[0007] In an exemplary embodiment, an initial vector conversion operation is performed on N of the above-mentioned node information to obtain N initial feature vectors, including: for each of the above-mentioned node information, an initial vector conversion operation is performed through the following steps to obtain N initial feature vectors: parsing the above-mentioned node information to determine the node type of the above-mentioned node; when the above-mentioned node type is the first node type, converting the static information included in the above-mentioned node information into a vector representation to obtain the above-mentioned initial feature vector; when the above-mentioned node type is the second node type, converting the dynamic timing information included in the above-mentioned node information into a vector representation to obtain the above-mentioned initial feature vector, wherein the above-mentioned dynamic timing information includes the timing relationship between the operation information and the time information generated by the above-mentioned node during the operation.

[0008] In an exemplary embodiment, a pre-trained first target network is used to perform a graph structure conversion operation on the N initial feature vectors to obtain a structural relationship graph between the N nodes, including: for each initial feature vector, determining the association relationship between the initial feature vector and other initial feature vectors, and setting the association relationship to the initial feature vector to obtain N target feature vectors; hierarchically fusing the N target feature vectors according to the node types of the N nodes to determine the structural relationship between the N nodes; and generating a structural relationship graph based on the structural relationship between the N nodes.

[0009] In an exemplary embodiment, for each of the above-mentioned initial feature vectors, the association relationship between the above-mentioned initial feature vector and other initial feature vectors is determined, and the above-mentioned association relationship is set in the above-mentioned initial feature vector to obtain N target feature vectors, including: obtaining a feature vector of query node information to obtain a query feature vector, wherein the above-mentioned query node information includes information of the user device querying the N above-mentioned nodes within a historical time period, and the above-mentioned query feature vector is used to represent the association relationship between the N above-mentioned nodes in the form of a vector; for each of the above-mentioned initial feature vectors, determining the above-mentioned association relationship between the above-mentioned other initial feature vectors from the above-mentioned query feature vector; setting the above-mentioned association relationship in the form of a vector to the above-mentioned initial feature vector to obtain N above-mentioned target feature vectors.

[0010] In an exemplary embodiment, according to the node types of the N nodes, the N target feature vectors are hierarchically fused to determine the structural relationship between the N nodes, including: determining M first nodes belonging to the first node type and K second nodes belonging to the second node type from the N target feature vectors, wherein both M and K are natural numbers less than N; constructing a first structural relationship between the M first nodes; constructing a second structural relationship between the K second nodes; according to the positional relationship and association relationship between the M first nodes and the K second nodes, the first structural relationship and the second structural relationship are fused to obtain the structural relationship between the N nodes.

[0011] In an exemplary embodiment, after the first structural relationship and the second structural relationship are integrated to obtain the structural relationship between the N nodes according to the positional relationship and the association relationship between the M first nodes and the K second nodes, the method further includes at least one of the following: inserting the query node into the structural relationship between the N nodes based on the query feature vector to obtain a third structural relationship, wherein the query feature vector is used to represent the association relationship between the N nodes in the form of a vector; inserting the signaling node into the structural relationship between the N nodes based on the signaling feature vector to obtain a fourth structural relationship, wherein the signaling feature vector is used to represent the signaling parameters of the signaling node in the form of a vector, and the signaling parameters include information generated by the signaling node in the process of transmitting data between the K second nodes.

[0012] In an exemplary embodiment, a structural relationship graph is generated based on the structural relationship between the N nodes, including: determining the node type of each node and the connection relationship between the nodes from the structural relationship between the N nodes; creating a node graph in a preset graphic tool according to the node type of each node; creating connection edges in the preset graphic tool according to the connection relationship between the nodes; and constructing a structural relationship graph based on the node graph and the connection edges.

[0013] In an exemplary embodiment, after using a pre-trained first target network to perform a graph structure conversion operation on the N initial feature vectors to obtain a structural relationship diagram between the N nodes, the method further includes: obtaining a target task sent by a target device, wherein the target task includes a current node where the target device is located and a target node to be reached; when the current node and the target node are found in the structural relationship diagram, a target path relationship diagram between the current node and the target node is generated in the structural relationship diagram.

[0014] In an exemplary embodiment, the first target network is a network obtained by iteratively training the first initial network using sample feature vectors of multiple sample nodes, wherein the first initial network includes a first feature extraction subnetwork and a feature fusion subnetwork, and for each of the sample feature vectors, the first feature extraction subnetwork is used to determine the sample association relationship between the sample feature vector and other sample feature vectors, and set the sample association relationship to the sample feature vector to obtain multiple sample target feature vectors; the feature fusion subnetwork is used to hierarchically fuse the multiple sample target feature vectors according to the sample node types of the multiple sample nodes to determine the sample structure relationship between the multiple sample nodes, and generate a sample structure relationship graph based on the sample structure relationship between the multiple sample nodes, wherein the sample structure relationship graph includes: the multiple sample nodes, the sample connection relationship between the multiple sample nodes, and the sample node types of the multiple sample nodes.

[0015] According to one embodiment of the present application, a method for generating a path relationship graph is provided, comprising: obtaining a target task sent by a target device, wherein the target task includes a current node where the target device is located and a target node to be traveled to; using a pre-trained second target network to output a target path relationship graph between the current node and the target node, wherein the target path relationship graph is a path relationship graph output by the second target network based on a structural relationship graph between N nodes output by the first target network, wherein the structural relationship graph includes: the N nodes, the connection relationships between the N nodes, and the node types of the N nodes, and the target path relationship graph includes the current node, the target node, the path between the current node and the target node, and the node types of the nodes included in the path.

[0016] In an exemplary embodiment, the second target network is a network obtained by iteratively training the second initial network using multiple sample tasks, wherein the second initial network includes a second feature extraction subnetwork and a classification subnetwork, and for each of the sample tasks, the second feature extraction subnetwork is used to extract node features of the current sample node and the target sample node to be reached in the sample task, and the sample task is a task for requesting to generate a sample path between the current sample node and the target sample node; the classification subnetwork is used to classify and identify the current sample node and the target sample node based on the node features of the current sample node and the node features of the target sample node, so as to classify the sample path from the sample structure relationship diagram.

[0017] According to another embodiment of the present application, a data processing system is provided, including: a first subsystem and a second subsystem, wherein the first subsystem is used to generate a data graph when executed, and the second subsystem is used to generate a path relationship graph when executed. According to another embodiment of the present application, a data graph generation device is provided, including: an acquisition module, used to acquire node information of N nodes in a target area, and obtain N node information, wherein the above N is a natural number greater than 1; a conversion module, used to perform an initial vector conversion operation on the N above node information, and obtain N initial feature vectors, wherein the above initial feature vector is used to represent the above node information in the form of a vector; a conversion module, used to perform a graph structure conversion operation on the N above initial feature vectors using a pre-trained first target network, and obtain a structural relationship graph between the N above nodes, wherein the above structural relationship graph includes: the N above nodes, the connection relationship between the N above nodes, and the node types of the N above nodes.

[0018] According to another embodiment of the present application, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the steps in any of the above method embodiments are implemented.

[0019] According to another embodiment of the present application, a computer-readable storage medium is provided, in which a computer program is stored, wherein the computer program is configured to execute the steps of any one of the above method embodiments when running.

[0020] According to another embodiment of the present application, an electronic device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute the steps in any one of the above method embodiments.

[0021] Through the present application, after obtaining the node information of N nodes in the target area, the obtained N node information, including dynamic timing information, etc., is converted into vectors to obtain N initial feature vectors, and then the N initial feature vectors are input into the pre-trained first target network. The target network fuses the N initial feature vectors according to the feature vector of the query node information and determines the relationship between the N nodes, and finally generates a structural relationship diagram. Thus, the structural relationship between each node can be accurately determined, and the structural relationship between each node can be displayed in the structural relationship diagram. Therefore, the problem that the structural relationship between multiple nodes cannot be accurately determined in the related art can be solved. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 It is a schematic diagram of the hardware environment of a data graph generation method according to an embodiment of the present application;

[0023] Figure 2 is a flow chart of a data graph generating method according to an embodiment of the present application;

[0024] Figure 3 is a schematic diagram of a POI node of a first node type according to an embodiment of the present application;

[0025] Figure 4 This is a schematic diagram of a POI node of the second node type according to an embodiment of the present application. Figure 1 ;

[0026] Figure 5 This is a schematic diagram of a POI node of the second node type according to an embodiment of the present application. Figure 2 ;

[0027] Figure 6 is a schematic diagram of a POI node according to an embodiment of the present application;

[0028] Figure 7 is a schematic diagram of the query node organization form according to an embodiment of the present application;

[0029] Figure 8 It is a schematic diagram of a structural relationship of inserting a query node into nodes based on a query feature vector according to an embodiment of the present application;

[0030] Fig. 9 According to the structural relationship of the embodiment of the present application Figure 1 ;

[0031] Fig.10 is a schematic diagram of a target task navigation path according to an embodiment of the present application;

[0032] Fig.11 A flow chart of a method for generating a path relationship graph according to an embodiment of the present application;

[0033] Fig.12 is a schematic diagram of an architecture for generating a target data graph according to a node of a first node type according to an embodiment of the present application;

[0034] Fig.13 This is a schematic diagram of the architecture of the node generating the target data graph according to the embodiment of the present application. Figure 1 ;

[0035] Fig.14 This is a schematic diagram of the architecture of the node generating the target data graph according to the embodiment of the present application. Figure 2 ;

[0036] Fig.15 is a structural block diagram of a data processing system according to an embodiment of the present application;

[0037] Fig.16It is a structural block diagram of a data graph generating device according to an embodiment of the present application. DETAILED DESCRIPTION

[0038] The embodiments of the present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0039] It should be noted that the terms "first", "second", etc. in the specification and claims of this application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

[0040] The method embodiments provided in the embodiments of the present application can be executed in a server device or a similar computing device. Taking running on a server device as an example, Figure 1 Schematic diagram of the hardware environment of a data graph generation method according to an embodiment of the present application. Figure 1 As shown, the server device may include one or more ( Figure 1 Only one is shown in the figure) a processor 102 (the processor 102 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA) and a memory 104 for storing data, wherein the server device may also include a transmission device 106 and an input / output device 108 for communication functions. It can be understood by those skilled in the art that Figure 1 The structure shown is only for illustration and does not limit the structure of the above server device. Figure 1 More or fewer components as shown, or with Figure 1 Different configurations are shown.

[0041] The memory 104 can be used to store computer programs, for example, software programs and modules of application software, such as a computer program corresponding to a data graph generation method in an embodiment of the present application. The processor 102 executes various functional applications and data processing by running the computer program stored in the memory 104, that is, to implement the above method. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include a memory remotely arranged relative to the processor 102, and these remote memories may be connected to a server device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0042] The transmission device 106 is used to receive or send data via a network. The specific example of the above network may include a wireless network provided by a communication provider of the server device. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, referred to as NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 106 can be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0043] In this embodiment, a data graph generation method is provided. Figure 2 is a flow chart of a data graph generation method according to an embodiment of the present application, such as Figure 2 As shown, the process includes the following steps:

[0044] Step S202, obtaining node information of N nodes in the target area, where N is a natural number greater than 1;

[0045] This embodiment includes but is not limited to being applied to scenarios where the association relationship between multiple nodes needs to be determined. For example, in a scenario of navigation on land or in sea, it is necessary to construct the position relationship between buildings, coordinate reference objects, and dynamically moving equipment to generate a real-time navigation path.

[0046] Optionally, the target area in this embodiment can be a certain area at sea, on land or in the air. In this embodiment, the value of N can be flexibly set based on the actual reference scene or the number of nodes required. For example, the target area includes 3 objects, which constitute 3 nodes. The 3 objects can be the same or different. The 3 objects include but are not limited to static objects and dynamic objects such as pedestrians, vehicles, and buildings.

[0047] Optionally, the node information in this embodiment may include multiple types of data in different application scenarios, aiming to comprehensively describe the characteristics of the node. The node information may include: geographic information (for example, the node's latitude and longitude coordinates, altitude, and detailed address); name; type (for example, static objects and dynamic objects); when the node type is a dynamic type, the node information may also include the object's moving direction, moving speed, node status, etc.

[0048] For example, in a scenario where the node information is communication, navigation, and telemetry fusion data, the data type of the node information and the data included therein are shown in Table 1.

[0049] Table 1:

[0050]

[0051] Among them, terminal signaling messages are sent by terminal users to broadcast real-time geographic information on land; Automatic Identification System message (AIS) and Automatic Dependent Surveillance-Broadcast message (ADS-B) are used in the fields of navigation and aviation respectively to broadcast the operating status and real-time longitude and latitude information of ships and aircraft; Internet of Things messages mainly involve various types of sensor data, such as monitoring data, equipment operation data and meteorological data.

[0052] Step S204, performing an initial vector conversion operation on the N node information to obtain N initial feature vectors, wherein the initial feature vectors are used to represent the node information in the form of vectors;

[0053] Optionally, in the scenario of processing graph data or network data, nodes and edges are involved. Node information can be any type of data, such as text description, numerical features, timestamps, etc. Initial vector conversion of node information is the process of converting this diverse information into a numerical vector that can be processed by machine learning algorithms.

[0054] For example, the initial vector conversion of node information includes the following schemes:

[0055] If the node information is text data, you can use methods such as Bag of Words, TF-IDF, Word2Vec, etc. to convert the text into a numerical vector. If the node information is an image, you can use a convolutional neural network (CNN) to extract features.

[0056] For node information of numerical type, the data can be scaled to the same scale through methods such as Min-Max Scaling and Z-score Standardization.

[0057] For categorical variables, one-hot encoding can be used to convert each category into a binary vector, with a 1 at each position indicating the presence of a category.

[0058] In deep learning, embedding layers can be used to learn low-dimensional representations in high-dimensional space. For example, in graph embedding techniques.

[0059] For some types of data, such as images or text, pre-trained models (such as BERT, ResNet, etc.) can be used to extract features.

[0060] Use machine learning algorithms (such as principal component analysis PCA and autoencoder) to automatically extract features.

[0061] The converted initial feature vector will be used to represent the node and can be used for tasks such as graph analysis, node classification, and link prediction.

[0062] Step S206, using the pre-trained first target network to perform a graph structure conversion operation on the N initial feature vectors to obtain a structural relationship graph between the N nodes, wherein the structural relationship graph includes: the N nodes, the connection relationships between the N nodes, and the node types of the N nodes.

[0063] Optionally, the first target network in this embodiment includes but is not limited to a machine learning model, such as a support vector machine (SVM), a random forest, a neural network, a Transformer-based graph neural network (GNN) architecture, etc. For another example, when the target network is a Transformer-based GNN architecture, the structural relationship diagram between N nodes can be expressed as: G = {L, E}, where L is a node set, E is an edge set, the node set can be each point of interest (POI for short), and the edge set describes the association method and degree of association between each POI node.

[0064] Through the above steps, after obtaining the node information of N nodes in the target area, the obtained N node information, including dynamic timing information, etc., is converted into vectors to obtain N initial feature vectors, and then the N initial feature vectors are input into the pre-trained first target network to generate a structural relationship diagram. Thus, the structural relationship between each node can be accurately determined, and the structural relationship between each node can be displayed in the structural relationship diagram. Therefore, the problem that the structural relationship between multiple nodes cannot be accurately determined in the related art can be solved.

[0065] In an exemplary embodiment, an initial vector conversion operation is performed on N of the above-mentioned node information to obtain N initial feature vectors, including: for each of the above-mentioned node information, an initial vector conversion operation is performed through the following steps to obtain N initial feature vectors: parsing the above-mentioned node information to determine the node type of the above-mentioned node; when the above-mentioned node type is the first node type, converting the static information included in the above-mentioned node information into a vector representation to obtain the above-mentioned initial feature vector; when the above-mentioned node type is the second node type, converting the dynamic timing information included in the above-mentioned node information into a vector representation to obtain the above-mentioned initial feature vector, wherein the above-mentioned dynamic timing information includes the timing relationship between the operation information and the time information generated by the above-mentioned node during the operation.

[0066] Optionally, the first node type in this embodiment includes but is not limited to a static node type, that is, the node information includes static information, and the static information includes but is not limited to the geographic data of the node and the time of obtaining the geographic data, wherein the geographic data includes monitoring data, equipment data, meteorological data and other IoT data. For example, when the node is a static POI node, the static POI node includes land buildings or identification points for land user navigation and route finding. Various sensors are installed inside the building for the purpose of monitoring, equipment detection, temperature and humidity statistics, etc. Therefore, the node information of the static POI node can be building POI geographic information data, including the name of the building, longitude and latitude coordinates, etc. Figure 3 The figure shows the data relationship between static POI nodes and edges, taking land buildings as an example. The map is constructed based on the overall distribution of POI nodes, and the edge weight is only determined by the physical distance between each POI node.

[0067] Optionally, the second node type in this embodiment includes but is not limited to a dynamic node type, that is, the node information includes dynamic information, and the dynamic information includes but is not limited to information about nodes that move during positioning and path finding. For example, for people using mobile phone navigation and path finding, ships at sea, aircraft in the air, etc., the geographic data included in the node information of the node of the second node type can be obtained in real time in combination with terminal signaling data, satellite messages, etc.

[0068] For example, Figure 4 As shown, in the land navigation scenario, the node type of the POI node is a dynamic node type, the 01-04 nodes represented by circles are all static node types, and the 05 node represented by a rectangle is a dynamic node type (for example, the 05 node is a pedestrian node). The node information of the 05 node includes basic data, geographic data and real-time data, and the real-time data includes but is not limited to motion data, physiological data and time. Motion data, physiological data, etc. can be collected by sensors such as sports bracelets carried by pedestrians.

[0069] For example, Figure 5 As shown, in the scenario of maritime navigation, node information can use AIS data and ADS-B data for communication and positioning. The 02 node represented by the circle is a node of the static node type, which can represent static objects such as buoys and lighthouses in the marine environment. The nodes represented by the rectangle are nodes of the dynamic node type, among which the 05 node is a sailing ship node. The node information of the 05 node can be obtained from devices such as gyroscopes, or from information parsed from AIS messages. The node information of the 05 node includes hull data, geographic data and real-time data, and the real-time data includes but is not limited to motion data, environmental data and time.

[0070] Optionally, the node information of the second type of node may also include data of a signaling node, for example, Figure 6 As shown, when the acquired nodes include the second type of nodes, T1 and T2 nodes are signaling nodes. The data of the signaling nodes have time series characteristics and can be used for historical motion trajectory analysis of POI nodes, current motion trend prediction, etc. The data of the signaling nodes are used for real-time positioning of dynamic POI points for subsequent generation of structural relationship diagrams. However, it should be noted that the data of the signaling nodes only plays a role in path auxiliary planning. When the data of the signaling nodes is missing, redundant, or the reliability is reduced due to base station drift, ping-pong effect, etc., or when the AIS and ADS-B signal transmitters are deliberately turned off when ships and aircraft perform specific tasks, causing the signaling data to be unavailable, it is only necessary to remove this signaling node, which will not affect the normal construction of other POI nodes and edges when the structural relationship diagram is generated later.

[0071] Optionally, in this embodiment, since the data of the signaling node only plays a role in assisting path planning, a "usage degree" indicator α can be introduced to describe the reference degree of the data of the signaling node. When the usage degree is 0, it means that the signaling data is not used at all. When the usage degree is 1, it means that the signaling data is completely reliable and can be used as a reference. The usage degree α∈[0, 1].

[0072] Optionally, in this embodiment, when the node type is the second node type, the process of converting the dynamic time series information included in the node information into a vector representation may be: based on the formula Calculate the timing code, where d total is the total number of N nodes in the target area, pos is the position of the second type of node, and t represents time.

[0073] The following is a specific example to illustrate the process of calculating timing encoding:

[0074] For example, when t=1, there is a vector of the form (0, 1), which means that there are nodes at both position 0 and position 1. total=2, at this time, the ROTE code of the node at position 0 is: The ROTE encoding of the node at position 1 is: The corresponding time series vector is

[0075]

[0076] Through the above steps, based on the node type, the specific information included in the node information is subjected to corresponding vector conversion operations to obtain the initial feature vector, thereby realizing the vector conversion operation of the node information and facilitating the subsequent graph structure conversion operation of the target network.

[0077] In an exemplary embodiment, a pre-trained first target network is used to perform a graph structure conversion operation on the N initial feature vectors to obtain a structural relationship graph between the N nodes, including: for each initial feature vector, determining the association relationship between the initial feature vector and other initial feature vectors, and setting the association relationship to the initial feature vector to obtain N target feature vectors; hierarchically fusing the N target feature vectors according to the node types of the N nodes to determine the structural relationship between the N nodes; and generating a structural relationship graph based on the structural relationship between the N nodes.

[0078] Optionally, the association relationship in this embodiment is used to indicate whether there is an association between the initial feature vectors of the nodes, including but not limited to whether the types of the nodes are similar.

[0079] Through the above steps, the first target network sets the association relationship between the current initial feature vector and other initial feature vectors in the current initial feature vector in the form of a vector, obtains the target feature vector, fuses it, determines the structural relationship between N nodes, and finally generates a structural relationship graph, determines the structural relationship between the nodes in a standardized form, and achieves the technical effect of improving the efficiency of generating the structural relationship graph.

[0080] In an exemplary embodiment, for each of the above-mentioned initial feature vectors, the association relationship between the above-mentioned initial feature vector and other initial feature vectors is determined, and the above-mentioned association relationship is set in the above-mentioned initial feature vector to obtain N target feature vectors, including: obtaining a feature vector of query node information to obtain a query feature vector, wherein the above-mentioned query node information includes information of the user device querying the N above-mentioned nodes within a historical time period, and the above-mentioned query feature vector is used to represent the association relationship between the N above-mentioned nodes in the form of a vector; for each of the above-mentioned initial feature vectors, determining the above-mentioned association relationship between the above-mentioned other initial feature vectors from the above-mentioned query feature vector; setting the above-mentioned association relationship in the form of a vector to the above-mentioned initial feature vector to obtain N above-mentioned target feature vectors.

[0081] Optionally, the query node information in this embodiment is the query log generated by the user for the location when the user previously queried. The query node information includes rich historical prior information, which can be used for edge type definition and edge weight calculation. Each piece of information in the query log is defined as a query node. The query node generally includes the user's target POI information. The target POI information mainly includes the name, address, type, longitude and latitude, etc. For example, Figure 7 As shown, the horizontal arrow represents the chronological order, and the nodes Q1 to Q4 above the horizontal arrow are query nodes, represented by squares. Each query node is stored in the order of the user's query, and the POI node directly below the query node represents the POI node where the user is located when querying.

[0082] Through the above steps, based on the information of N nodes queried by the user device in the historical time period, the association relationship between the N nodes can be determined, which is converted into a vector form and set in the initial feature vector of the N nodes, and the final N target feature vectors are further determined to facilitate the subsequent generation of the corresponding structural relationship diagram and display the relationship between each node.

[0083] In an exemplary embodiment, according to the node types of the N nodes, the N target feature vectors are hierarchically fused to determine the structural relationship between the N nodes, including: determining M first nodes belonging to the first node type and K second nodes belonging to the second node type from the N target feature vectors, wherein both M and K are natural numbers less than N; constructing a first structural relationship between the M first nodes; constructing a second structural relationship between the K second nodes; according to the positional relationship and association relationship between the M first nodes and the K second nodes, the first structural relationship and the second structural relationship are fused to obtain the structural relationship between the N nodes.

[0084] Optionally, the first structural relationship in this embodiment represents an association relationship between M first nodes of the first node type, and the first structural relationship includes the correlation between the nodes and the similarity of the nodes under geographical clustering conditions. The value of M can be determined based on actual scenarios.

[0085] Optionally, the structural relationship between the N nodes in this embodiment includes the location correlation, functional correlation and similarity between different POI nodes. Among them, the correlation between POI nodes in geographical locations can be determined based on the distance between them. For example, two adjacent stores may have a higher location correlation due to their close geographical locations. The functional correlation of POI nodes refers to whether the roles they play or the services they provide in a specific context are similar or complementary. For example, a cafe and a bookstore may have functional correlation because they both provide a relaxing environment. The similarity of POI nodes can be measured based on their characteristics or attributes. For example, two restaurants that both provide vegetarian food may have a high similarity in the type of dishes.

[0086] Through the above steps, the structural relationship between various types of nodes is determined, and thus the structural relationship between N nodes is determined, so as to facilitate the subsequent generation of the corresponding structural relationship diagram and display the relationship between each node.

[0087] In an exemplary embodiment, after the first structural relationship and the second structural relationship are integrated to obtain the structural relationship between the N nodes according to the positional relationship and the association relationship between the M first nodes and the K second nodes, the method further includes at least one of the following: inserting the query node into the structural relationship between the N nodes based on the query feature vector to obtain a third structural relationship, wherein the query feature vector of the query node is used to represent the association relationship between the N nodes in the form of a vector; inserting the signaling node into the structural relationship between the N nodes based on the signaling feature vector to obtain a fourth structural relationship, wherein the signaling feature vector is used to represent the signaling parameters of the signaling node in the form of a vector, and the signaling parameters include information generated by the signaling node in the process of transmitting data between the K second nodes.

[0088] Optionally, in this embodiment, the query node is inserted into the structural relationship between these nodes. This may involve the operation of a graph database, where nodes and edges represent entities and the relationships between them. For example, Figure 8 As shown, the query node can be inserted into the structural relationship between the six nodes based on the query feature vector obtained from the query log to obtain a third structural relationship. The specific insertion method includes: determining the node related to the query node based on the node information included in the query feature vector, associating the query node with the node, and obtaining the third structural relationship.

[0089] Optionally, in this embodiment, the signaling node is inserted into the structural relationship between N nodes based on the signaling feature vector to obtain the fourth structural relationship. The specific insertion method includes: based on the signaling feature vector, determining two nodes related to the signaling node, inserting the signaling node into the two nodes, and obtaining the fourth structural relationship.

[0090] Optionally, the signaling parameters in this embodiment include information generated during data transmission between the signaling node and the second node, or information generated during data transmission between the first nodes obtained based on the query node, and the information generated during data transmission may be node status information. Figure 6 As shown, the T1 signaling node is the status information sent by the dynamic POI node 07 during navigation and path finding, and the T2 signaling node is the status information sent by other terminals.

[0091] Through the above steps, the query node and the signaling node are inserted into the structural relationship of N nodes based on the query feature vector and the signaling feature vector, which enriches the structural relationship between the N nodes and more comprehensively represents the structural relationship between the N nodes, facilitating the subsequent generation of a structural relationship diagram.

[0092] In an exemplary embodiment, a structural relationship graph is generated based on the structural relationship between the N nodes, including: determining the node type of each node and the connection relationship between the nodes from the structural relationship between the N nodes; creating a node graph in a preset graphic tool according to the node type of each node; creating connection edges in the preset graphic tool according to the connection relationship between the nodes; and constructing a structural relationship graph based on the node graph and the connection edges.

[0093] Optionally, the node graph in this embodiment is displayed according to different node types.

[0094] Optionally, the connection edges in this embodiment include "query" edges, "start and end" edges and "close to" edges, and the connection of the edges represents the correlation between different nodes. Among them, the "query" edge describes the POI search information at a specific location; the "start and end" edge describes the actual navigation start and end point positions after receiving the target task sent by the target device. In the land scenario, it can be the start and end positions in the navigation process. In the ocean and aerospace fields, it can be the actual geographical location of various cooperative targets and non-cooperative targets; the difference between the "query" edge and the "start and end" edge is that the "query" edge only performs queries (for example, searching for several suitable restaurants in a certain area), and is not necessarily used to generate navigation and route planning based on the target task (for example, although several restaurants are queried, only one is selected for dining); the "close to" edge describes the similarity of POI points under geographical agglomeration conditions. Since the geographical area division has certain similarities, other environments near the POI may also be points of interest corresponding to the target task.

[0095] For example, Fig. 9 As shown in the figure, the thickness of the edge represents the edge weight. The edge between node 01 and node 02 is thickened to indicate that the strength of the association between the two is greater than the strength of the association between the two and other nodes. In the "query" edge, each query node is connected to the POI node at its location in the form of a first dotted line; in the "similar" edge, each POI node is connected in the form of a second dotted line.

[0096] Through the above steps, the node type of each node and the connection relationship between each node are determined from the structural relationship between N nodes, node graphs and connection edges are created, and a structural relationship diagram is constructed, which helps to clearly display the relationship between each node and better understand and analyze the entire structure.

[0097] In an exemplary embodiment, after using a pre-trained first target network to perform a graph structure conversion operation on the N initial feature vectors to obtain a structural relationship diagram between the N nodes, the method further includes: obtaining a target task sent by a target device, wherein the target task includes a current node where the target device is located and a target node to be reached; when the current node and the target node are found in the structural relationship diagram, a target path relationship diagram between the current node and the target node is generated in the structural relationship diagram.

[0098] Optionally, the target task in this embodiment includes but is not limited to the task of performing node path planning. Fig.10 As shown, the current node where the target device is located and the target node to be reached are node 01 and node 06 respectively, the target task navigation path is node 01-node 03-node 02-node 06, and the final structural relationship diagram includes the target navigation path.

[0099] Through the above steps, on the basis of the generated structural relationship diagram, the navigation path is obtained according to the target task, the navigation path is added to the structural relationship diagram, and the target path relationship diagram is generated, so that the target path relationship diagram can achieve the effect of displaying the association relationship between nodes on the basis of obtaining the navigation path corresponding to the target task.

[0100] In an exemplary embodiment, the first target network is a network obtained by iteratively training the first initial network using sample feature vectors of multiple sample nodes, wherein the first initial network includes a first feature extraction subnetwork and a feature fusion subnetwork, and for each of the sample feature vectors, the first feature extraction subnetwork is used to determine the sample association relationship between the sample feature vector and other sample feature vectors, and set the sample association relationship to the sample feature vector to obtain multiple sample target feature vectors; the feature fusion subnetwork is used to hierarchically fuse the multiple sample target feature vectors according to the sample node types of the multiple sample nodes to determine the sample structure relationship between the multiple sample nodes, and generate a sample structure relationship graph based on the sample structure relationship between the multiple sample nodes, wherein the sample structure relationship graph includes: the multiple sample nodes, the sample connection relationship between the multiple sample nodes, and the sample node types of the multiple sample nodes.

[0101] Optionally, the sample association relationship in this embodiment is used to represent the association relationship between the current sample feature vector and other sample feature vectors in the form of a vector, including but not limited to: the samples belong to the same type, the samples have similar functions, etc.

[0102] Optionally, the sample node types in this embodiment include but are not limited to: static nodes, dynamic nodes, signaling nodes, etc.

[0103] Optionally, the first initial network in this embodiment includes but is not limited to a machine learning model, such as a support vector machine (SVM), a random forest, a neural network, a Transformer-based graph neural network (GNN) architecture, etc. For another example, when the first initial network is a Transformer-based GNN architecture, the sample structure relationship graph between samples can be expressed as: G = {L, E}, where L is a node set, E is an edge set, the node set can be each point of interest (POI for short), and the edge set describes the association method and degree of association between each POI node.

[0104] Optionally, the first target network in this embodiment is a network obtained by iteratively training the first initial network using sample feature vectors of multiple sample nodes. When the sample structure relationship graph generated by the trained first initial network meets the needs of operators in the field or the degree of determination of the structural relationship between individual samples in the sample structure relationship graph generated by the trained first initial network meets a threshold, the iterative training of the first initial network is stopped to obtain the first target network.

[0105] Through the above steps, the first initial network is iteratively trained using the sample feature vectors of multiple sample nodes. After determining the sample association relationship between the sample feature vector and other sample feature vectors, the first initial network obtains multiple sample target feature vectors, and based on this, feature fusion is performed to further generate a sample structure relationship diagram. The first initial network can be iteratively trained according to the sample feature vectors of the sample nodes, and the final first target network can be determined according to actual conditions, which is convenient for the subsequent generation of a structure relationship diagram between N nodes.

[0106] In this embodiment, a method for generating a path relationship graph is provided. Fig.11 is a flow chart of a method for generating a path relationship graph according to an embodiment of the present application. Fig.11 As shown, the process includes the following steps:

[0107] Step S1102, obtaining a target task sent by a target device, wherein the target task includes a current node where the target device is located and a target node to be traveled to;

[0108] This embodiment includes but is not limited to being applied to scenarios where the current node and the target node need to be determined. For example, in a land navigation scenario, a pedestrian or vehicle needs to generate a navigation path to reach a target building or target location; or a ship in the sea needs to navigate to a target island or meet a target ship.

[0109] Optionally, the target device in this embodiment is a device that can send information, including but not limited to: a mobile phone, a tablet, a car navigation system, a ship, a smart watch, etc.

[0110] Optionally, the target task in this embodiment is a navigation task, including the current position sent by the target device and the target node to be reached, wherein the target node can be a static node or a dynamic node.

[0111] Step S1104, use the pre-trained second target network to output the target path relationship graph between the above-mentioned current node and the above-mentioned target node, wherein the above-mentioned target path relationship graph is a path relationship graph output by the above-mentioned second target network based on the structural relationship graph between the N nodes output by the first target network, and the above-mentioned structural relationship graph includes: the N above-mentioned nodes, the connection relationship between the N above-mentioned nodes, and the node types of the N above-mentioned nodes, and the above-mentioned target path relationship graph includes the above-mentioned current node, the above-mentioned target node, the path between the above-mentioned current node and the above-mentioned target node, and the node types of the nodes included in the above-mentioned path.

[0112] Optionally, the second target network in this embodiment includes but is not limited to a machine learning model, such as a support vector machine (SVM), a random forest, a neural network, a Transformer-based graph neural network (GNN) architecture, etc.

[0113] Optionally, the target path relationship graph in this embodiment includes the current node, the target node, the path between the current node and the target node, the node types of the nodes included in the path, and may also include: other nodes existing in the path between the current node and the target node, the node names and node types of other nodes, and the structural relationships between other nodes in the path.

[0114] Through the above steps, after obtaining the target task sent by the target device, the second target network can output a path relationship diagram based on the structure relationship diagram, thereby improving the efficiency of obtaining the target path and being able to quickly generate the target path relationship diagram.

[0115] In an exemplary embodiment, the second target network is a network obtained by iteratively training the second initial network using multiple sample tasks, wherein the second initial network includes a second feature extraction subnetwork and a classification subnetwork, and for each of the sample tasks, the second feature extraction subnetwork is used to extract node features of the current sample node and the target sample node to be reached in the sample task, and the sample task is a task for requesting to generate a sample path between the current sample node and the target sample node; the classification subnetwork is used to classify and identify the current sample node and the target sample node based on the node features of the current sample node and the node features of the target sample node, so as to classify the sample path from the sample structure relationship diagram.

[0116] Optionally, the second target network in this embodiment has been described above and will not be described again here.

[0117] Optionally, the sample task in this embodiment is used to represent going from node 1 to a target node, including the current node and the target node to be traveled to.

[0118] Optionally, the sample path in this embodiment is a path determined by the second initial network according to the characteristics of the current sample node, the target sample node and the connection relationship between them. During the training process, the second initial network gradually optimizes its parameters by learning the correct generation of the sample path, thereby improving the accuracy and efficiency of path prediction.

[0119] Optionally, the second target network in this embodiment is a network obtained by iteratively training the second initial network using multiple sample tasks. When the sample path generated by the trained second initial network meets preset conditions, the iterative training of the second initial network is stopped to obtain the second target network, wherein the preset conditions may be that the time corresponding to the path is the shortest, the obstacles in the path are the least, etc.

[0120] Through the above steps, the second initial network is iteratively trained using multiple sample tasks, features are extracted for each sample task, and then the current sample nodes and the target sample nodes are classified and identified to obtain sample paths. The final second target network can be obtained by continuously iteratively training the second initial network, thereby improving the efficiency and accuracy of the path generated by the second target network.

[0121] The data receiving method in the embodiment of the present application is described in detail below with reference to optional examples.

[0122] This embodiment takes the fusion of different types of node information based on a graph neural network architecture as an example to illustrate.

[0123] The graph neural network architecture in this embodiment is a Transformer-based network architecture, which includes three layers of networks: the first layer of the network and the second layer of the network are equivalent to the first target network mentioned above, the third layer of the network is equivalent to the second target network mentioned above, the first layer of the network is equivalent to the first feature extraction subnetwork mentioned above, and the second layer of the network is equivalent to the feature fusion subnetwork mentioned above: the first layer of the network includes multiple subnetworks. After the initial feature vector of each node information is input into the first layer of the network, the first layer of the network first processes the initial feature vector using different subnetworks according to the different types of nodes, and determines each initial feature vector based on the information included in the initial feature vectors between the nodes and the information included in the query node and the signaling node. The target feature vector of each node is outputted based on the correlation between the feature vector and other initial feature vectors; then, the target feature vector of each node output by the first layer network is used as input to the second layer network, and the second layer network determines the structural relationship between nodes of each type based on the type of node, and inserts the query node into the structural relationship between nodes based on the feature vectors of the query node and the signaling node, creates nodes and creates edges based on the structural relationship of each node, and obtains a structural relationship graph; the third layer network generates paths between nodes based on the target task, and finally obtains a target path relationship graph corresponding to the target task, wherein the third layer network includes multiple third layer sub-networks, which can process multiple target tasks at the same time. The third layer network can fine-tune its network parameters according to the characteristics of the target task, and freeze all parameters of the first layer network and the second layer network during fine-tuning, and only train the third layer network.

[0124] For example, when the acquired node information only includes static node information, the architecture of the generated target data graph is as follows: Fig.12 As shown, node 01, node 03, node 02, and node 06 are all static type nodes. The initial feature vectors corresponding to node 01, node 03, node 02, node 06, and Q1 query node are input, and the first layer Transformer (equivalent to the first feature extraction subnetwork in the first target network) determines the association relationship between each initial feature vector and other initial feature vectors based on the information included in the initial feature vectors between static nodes and the information included in the query node, processes the initial feature vectors, and obtains the type of each node and the target feature vector, h cls Represents the node category displayed in vector form ( Fig.12 where are query nodes and nodes of the first node type, i.e., static nodes), H context It represents the feature information extracted by the Transformer architecture in the form of a vector. The number of node information and the number of transformers are not one-to-one. Fig.12The form shown is only for the convenience of understanding the processing process of the initial feature vector of the node; then the target feature vector is input into the graph Transformer integration layer, which (equivalent to the feature fusion subnetwork in the first target network mentioned above) will retain the association information of the POI nodes "01 node-03 node-02 node-06 node" and the POI node sequence information of the entire graph, and determine the structural relationship between each static node according to the type of node, and insert the query node into the structural relationship between the nodes according to the query node feature vector, obtain the structural relationship between each node, and output the structural relationship graph, wherein the above information is not for a specific target task, but only implicitly represents the degree of association of each POI node, historical path planning scheme and other information; finally, the Fine-tune layer (equivalent to the second target network mentioned above) creates nodes and creates edges according to the structural relationship of each node, generates paths between nodes according to the target task, and finally obtains the target path relationship graph corresponding to the target task.

[0125] Among them, for specific target tasks, several Fine-tune layers are added to fine-tune the structural relationship graph, where each Fine-tune layer corresponds to a target task. During fine-tuning, all parameters below the Transformer integration layer of the graph are frozen, and only the last Fine-tune layer is trained. The task result output by the final Fine-tune layer is a target path relationship graph containing a navigation path based on the target task.

[0126] Through the above embodiment, the feature vectors of static nodes and query nodes are processed, and based on this, they are converted into a target structure diagram, and a target schematic diagram containing a navigation path is generated according to the target task, so that the association relationship between nodes can be obtained while obtaining the navigation path.

[0127] For another example, when the acquired node information includes static node information and dynamic node information, the architecture of the generated target data graph is as follows: Fig.13As shown, nodes 01 and 03 are static nodes, and nodes 02 and 06 are dynamic nodes. The initial feature vectors corresponding to nodes 01, 03, 02, 06, and the Q1 query node are input, and time t is introduced at the same time. After a fixed time interval, the information of nodes 02 and 06 is collected and input into the Transformer layer (equivalent to the first feature extraction subnetwork in the above-mentioned first target network) for feature update. The Transformer of the first layer (equivalent to the first feature extraction subnetwork in the above-mentioned first target network) determines the association relationship between each initial feature vector and other initial feature vectors based on the information included in the initial feature vectors between nodes and the information included in the query node, and inserts time t into the feature vector of the dynamic node to obtain the type and target feature vector of each node, h cls Represents the node category displayed in vector form ( Fig.13 where the query node is a node of the first node type, i.e., a static node, and a node of the second node type, i.e., a dynamic node). context It represents the feature information extracted by the Transformer architecture in the form of a vector. The number of node information and the number of Transformers are not one-to-one. Fig.13 The form shown is only for the convenience of understanding the processing of the initial feature vector of the node; then it is input into the graph Transformer integration layer (equivalent to the feature fusion subnetwork in the first target network mentioned above), and the structural relationship between each node is determined according to the type of node, and the query node is inserted into the structural relationship between the nodes according to the query node feature vector, and the structural relationship between each node is obtained, and the structural relationship diagram is output. The subsequent Fine-tune layer (equivalent to the second target network mentioned above) is operated with Fig.12 The processing flow of the architecture diagram shown is similar and will not be repeated here.

[0128] Through the above embodiment, the feature vectors of static nodes, query nodes and dynamic nodes are processed, and based on this, they are converted into a target structure diagram, and a target path diagram containing a navigation path is generated according to the target task, so that the association relationship between nodes can be obtained while obtaining the navigation path and displaying the real-time position of the dynamic nodes.

[0129] For another example, when the acquired node information includes static node information, dynamic node information, and signaling node information, the architecture of the target data graph is generated as follows: Fig.14As shown, the 01 node and the 03 node are static type nodes, the 02 node and the 06 node are dynamic type nodes, and T1 is a signaling node. The initial feature vectors corresponding to the 01 node, the 03 node, the 02 node, the 06 node, the Q1 query node, and the T1 signaling node are input, and the time t and the signaling node and its usage α are introduced. After a fixed time interval, the information of the 02 and 06 nodes is collected and input into the Transformer layer (equivalent to the first feature extraction subnetwork in the above-mentioned first target network) for feature update. At the same time, based on the usage of the signaling node, it is determined whether the information represented by the signaling node is available. The Transformer of the first layer (equivalent to the first feature extraction subnetwork in the above-mentioned first target network) determines the association relationship between each initial feature vector and other initial feature vectors based on the information included in the initial feature vector between the nodes and the information included in the query node. When the signaling node is available, the association relationship between each initial feature vector and other initial feature vectors is determined in combination with the information included in the initial feature vector of the signaling node, and the time t is inserted into the feature vector of the dynamic node to obtain the type of the node and the target feature vector, h cls Represents the node category displayed in vector form ( Fig.14 where H is a query node, a signaling node, a node of the first node type, i.e., a static node, and a node of the second node type, i.e., a dynamic node). context It represents the feature information extracted by the Transformer architecture in the form of a vector. The number of node information and the number of Transformers are not one-to-one. Fig.14 The form shown is only for the convenience of understanding the processing process of the initial feature vector of the node; then it is input into the graph Transformer integration layer (equivalent to the feature fusion subnetwork in the first target network mentioned above), and the structural relationship between each node is determined according to the type of node, and the query node is inserted into the structural relationship between the nodes according to the query node feature vector, and the signaling node is inserted into the structural relationship between the nodes according to the signaling node feature vector, and the structural relationship between each node is obtained, and the structural relationship diagram is output; the subsequent Fine-tune layer (equivalent to the second target network mentioned above) is operated and Fig.12 The processing flow of the architecture diagram shown is similar and will not be repeated here.

[0130] Through the above embodiments, the feature vectors of various types of nodes are processed and converted into a target structure diagram based on this, and a target path diagram containing a navigation path is generated according to the target task. This achieves the effect of obtaining the association relationship between nodes while obtaining the navigation path, and further enriching the real-time position of the dynamic nodes displayed according to the signaling data.

[0131] It should be noted that, through the description of the above implementation methods, those skilled in the art can clearly understand that the method according to the above embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, a disk, or an optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods described in each embodiment of the present application.

[0132] Fig.15 is a structural block diagram of a data processing system according to an embodiment of the present application, such as Fig.15 As shown, the system includes:

[0133] The first subsystem 152 is used to implement the steps of the method in the above-mentioned data graph generation method when executed; including a first target network 15202;

[0134] The second subsystem 154 is used to implement the steps of the above-mentioned path relationship graph generation method when executed; including a second target network 15402. In this embodiment, a data graph generation device is also provided, which is used to implement the above-mentioned embodiments and preferred implementation modes, and will not be repeated hereafter. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.

[0135] Fig.16 is a structural block diagram of a data graph generating device according to an embodiment of the present application, such as Fig.16 As shown, the device comprises:

[0136] The first memory 162, the first processor 164, and the first computer program 16202 stored in the first memory 162 and executable on the first processor 164, the first processor 162 performs the following operations when executing the first computer program 16202: obtaining node information of N nodes in the target area, wherein N is a natural number greater than 1; performing an initial vector conversion operation on the N node information to obtain N initial feature vectors, wherein the initial feature vector is used to represent the node information in the form of a vector; using a pre-trained target network to perform a graph structure conversion operation on the N initial feature vectors to obtain a structural relationship graph between the N nodes, wherein the graph structure conversion operation includes: for each initial feature vector, determining an association relationship between the initial feature vector and other initial feature vectors, and setting the association relationship to the initial feature vector to obtain N target feature vectors; hierarchically fusing the N target feature vectors according to the node types of the N nodes to determine the structural relationship between the N nodes; and generating the structural relationship graph based on the structural relationship between the N nodes.

[0137] It should be noted that the above modules can be implemented by software or hardware. For the latter, it can be implemented in the following ways, but not limited to: the above modules are all located in the same processor; or the above modules are located in different processors in any combination.

[0138] An embodiment of the present application further provides a computer-readable storage medium, in which a computer program is stored, wherein the computer program is configured to execute the steps of any of the above method embodiments when running.

[0139] In an exemplary embodiment, the computer-readable storage medium may include, but is not limited to, various media that can store computer programs, such as a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk.

[0140] An embodiment of the present application further provides an electronic device, including a memory and a processor, wherein a computer program is stored in the memory, and the processor is configured to run the computer program to execute the steps in any one of the above method embodiments.

[0141] In an exemplary embodiment, the electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the processor, and the input / output device is connected to the processor.

[0142] An embodiment of the present application further provides a computer program product, which includes a computer program. When the computer program is executed by a processor, the steps in any one of the above method embodiments are implemented.

[0143] An embodiment of the present application further provides another computer program product, including a non-volatile computer-readable storage medium, wherein the non-volatile computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps in any of the above method embodiments are implemented.

[0144] The embodiments of the present application also provide a computer program, which includes computer instructions, which are stored in a computer-readable storage medium; a processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device performs the steps of any one of the above method embodiments.

[0145] For specific examples in this embodiment, reference may be made to the examples described in the above embodiments and exemplary implementation modes, and this embodiment will not be described in detail herein.

[0146] Obviously, those skilled in the art should understand that the above modules or steps of the present application can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, and in some cases, the steps shown or described can be executed in a different order from that herein, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present application is not limited to any specific combination of hardware and software.

[0147] The above description is only the preferred embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the principles of the present application shall be included in the protection scope of the present application.

Claims

1. A data graph generation method, characterized in that: include: Obtain node information of N nodes in the target area, where N is a natural number greater than 1; Performing an initial vector conversion operation on the N node information to obtain N initial feature vectors, wherein the initial feature vectors are used to represent the node information in the form of vectors; Use a pre-trained first target network to perform a graph structure conversion operation on the N initial feature vectors to obtain a structural relationship graph between the N nodes, wherein the structural relationship graph includes: the N nodes, the connection relationships between the N nodes, and the node types of the N nodes.

2. The method according to claim 1, characterized in that Performing an initial vector conversion operation on the N node information to obtain N initial feature vectors, including: For each of the node information, the initial vector conversion operation is performed through the following steps to obtain N initial feature vectors: Parsing the node information to determine the node type of the node; In the case where the node type is the first node type, converting the static information included in the node information into a vector representation to obtain the initial feature vector; In the case where the node type is the second node type, the dynamic timing information included in the node information is converted into a vector representation to obtain the initial feature vector, wherein the dynamic timing information includes the timing relationship between the operation information and time information generated by the node during the operation.

3. The method according to claim 1, characterized in that Using the pre-trained first target network to perform a graph structure conversion operation on the N initial feature vectors, to obtain a structural relationship graph between the N nodes, including: For each of the initial feature vectors, determining an association relationship between the initial feature vector and other initial feature vectors, and setting the association relationship to the initial feature vector to obtain N target feature vectors; According to the node types of the N nodes, the N target feature vectors are layered and fused to determine the structural relationship between the N nodes; A structural relationship graph is generated based on the structural relationship between the N nodes.

4. The method according to claim 3, characterized in that For each of the initial feature vectors, determining the association relationship between the initial feature vector and other initial feature vectors, and setting the association relationship to the initial feature vector, to obtain N target feature vectors, including: Acquire a feature vector of query node information to obtain a query feature vector, wherein the query node information includes information of N nodes queried by a user device in a historical time period, and the query feature vector is used to represent an association relationship between the N nodes in the form of a vector; For each of the initial feature vectors, determining the association relationship between the query feature vector and the other initial feature vectors; The association relationship is set in the initial feature vector in the form of a vector to obtain N target feature vectors.

5. The method according to claim 3, characterized in that: According to the node types of the N nodes, the N target feature vectors are hierarchically fused to determine the structural relationship between the N nodes, including: Determine M first nodes belonging to a first node type and K second nodes belonging to a second node type from the N target feature vectors, wherein both M and K are natural numbers less than N; Constructing a first structural relationship between the M first nodes; Constructing a second structural relationship between K of the second nodes; According to the positional relationship and the association relationship between the M first nodes and the K second nodes, the first structural relationship and the second structural relationship are fused to obtain the structural relationship between the N nodes.

6. The method according to claim 5, characterized in that After fusing the first structural relationship and the second structural relationship according to the positional relationship and the association relationship between the M first nodes and the K second nodes to obtain the structural relationship between the N nodes, the method further includes at least one of the following: Inserting the query node into the structural relationship between the N nodes based on the query feature vector to obtain a third structural relationship, wherein the query feature vector is used to represent the association relationship between the N nodes in the form of a vector; Based on the signaling feature vector, the signaling node is inserted into the structural relationship between N of the nodes to obtain a fourth structural relationship, wherein the signaling feature vector is used to represent the signaling parameters of the signaling node in the form of a vector, and the signaling parameters include information generated by the signaling node in the process of transmitting data between K of the second nodes.

7. The method according to claim 3, characterized in that Generating a structural relationship graph based on the structural relationship between the N nodes includes: Determining the node type of each of the nodes and the connection relationship between the nodes from the structural relationship between the N nodes; Creating a node graph in a preset graph tool according to the node type of each of the nodes; Creating connection edges in the preset graphic tool according to the connection relationship between each of the nodes; A structural relationship graph is constructed based on the node graph and the connection edges.

8. The method according to claim 1, characterized in that After performing a graph structure conversion operation on the N initial feature vectors using the pre-trained first target network to obtain a structural relationship graph between the N nodes, the method further includes: Acquire a target task sent by a target device, wherein the target task includes a current node where the target device is located and a target node to be traveled to; In the case where the current node and the target node are found in the structural relationship graph, a target path relationship graph between the current node and the target node is generated in the structural relationship graph.

9. The method according to claim 1, characterized in that: The first target network is a network obtained by iteratively training the first initial network using sample feature vectors of multiple sample nodes, wherein: The first initial network includes a first feature extraction subnetwork and a feature fusion subnetwork. For each of the sample feature vectors, the first feature extraction subnetwork is used to determine a sample association relationship between the sample feature vector and other sample feature vectors, and set the sample association relationship to the sample feature vector to obtain multiple sample target feature vectors; The feature fusion subnetwork is used to hierarchically fuse multiple sample target feature vectors according to the sample node types of the multiple sample nodes to determine the sample structure relationship between the multiple sample nodes, and generate a sample structure relationship graph based on the sample structure relationship between the multiple sample nodes, wherein the sample structure relationship graph includes: multiple sample nodes, sample connection relationships between the multiple sample nodes, and sample node types of the multiple sample nodes.

10. A method for generating a path relationship graph, characterized in that: include: Acquire a target task sent by a target device, wherein the target task includes a current node where the target device is located and a target node to be traveled to; A pre-trained second target network is used to output a target path relationship graph between the current node and the target node, wherein the target path relationship graph is a path relationship graph output by the second target network based on a structural relationship graph between N nodes output by the first target network, wherein the structural relationship graph includes: N nodes, connection relationships between the N nodes, and node types of the N nodes, and the target path relationship graph includes the current node, the target node, a path between the current node and the target node, and node types of the nodes included in the path.

11. The method according to claim 10, characterized in that The second target network is a network obtained by iteratively training the second initial network using multiple sample tasks, wherein: The second initial network includes a second feature extraction subnetwork and a classification subnetwork. For each of the sample tasks, the second feature extraction subnetwork is used to extract node features of a current sample node and a target sample node to be reached in the sample task. The sample task is a task for requesting generation of a sample path between the current sample node and the target sample node. The classification subnetwork is used to classify and identify the current sample node and the target sample node based on the node features of the current sample node and the node features of the target sample node, so as to classify the sample path from the sample structure relationship diagram.

12. A data processing system, characterized in that: include: A first subsystem and a second subsystem, wherein the first subsystem is used to implement the steps of the method described in any one of claims 1 to 9 when executed, and the second subsystem is used to implement the steps of the method described in claim 10 or 11 when executed.

13. 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 described in any one of claims 1 to 9 are implemented; or, when the computer program is executed by a processor, the steps of the method described in claim 10 or 11 are implemented.

14. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, wherein when the computer program is executed by a processor, the steps of the method described in any one of claims 1 to 9 are implemented, or when the computer program is executed by a processor, the steps of the method described in claim 10 or 11 are implemented.

15. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the steps of the method described in any one of claims 1 to 9 are implemented; or, when the computer program is executed by the processor, the steps of the method described in claim 10 or 11 are implemented.

Images