User WEB data analysis method and device based on graph neural network, equipment and medium
The user WEB data analysis method using graph neural networks improves product recommendation and behavioral safety predictions by integrating historical and current data through graph structures and autoencoders, addressing the inaccuracies and security issues in existing systems.
Patent Information
- Application Number
- CN202411184255.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-27
- Publication Date
- 2024-12-13
AI Technical Summary
Existing user interest prediction and behavior security systems in web shopping are inaccurate due to reliance on historical data that becomes unreliable and fail to consider correlations between user interests and behaviors, leading to ineffective recommendations and security issues.
A user WEB data analysis method using graph neural networks that incorporates historical and current data, generates graph structures with additional nodes to transmit unconnected information, and employs graph autoencoders to predict user interests and behavior safety, improving recommendation accuracy and security prediction.
Enhances product recommendation accuracy based on user interests and improves behavioral safety predictions by integrating graph neural networks and autoencoders, addressing the limitations of existing one-way prediction systems.
Smart Images
Figure CN119128231A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of computer artificial intelligence technology, and in particular to a user WEB data analysis method, device, equipment and medium based on graph neural network. Background Art
[0002] When people shop on the web, they typically click, purchase, and select products and conversations based on their personal interests and needs. Existing technologies typically use graph databases or graph structures to represent this scenario. Based on the user's historical interests and needs, graph databases or graph structures are used to predict the links, products, and conversations that the user is currently most interested in and make recommendations.
[0003] At the same time, as existing artificial intelligence technology becomes more and more advanced, the historical data of many links, products and conversations are no longer reliable, which will lead to the failure of predictions about user interests and interference between the same or similar users. At the same time, this rampant artificial intelligence behavior will also trigger corresponding network security incidents.
[0004] The existing technologies for predicting user interests and user behaviors are one-way, and do not consider the correlation between the prediction of user interests and the prediction of user behaviors. As a result, it is impossible to effectively correlate user interests and user behavior security, resulting in the inability to effectively predict user interests and user behavior security. Summary of the Invention
[0005] The main purpose of the embodiments of the present invention is to propose a user WEB data analysis method, device, equipment and medium based on graph neural network, which improves the accuracy of recommending products of interest to users and improves the accuracy of predicting behavioral safety when purchasing products.
[0006] One aspect of the present invention provides a user WEB data analysis method based on a graph neural network, comprising:
[0007] Obtain the user's historical WEB data and current WEB data, initialize the historical WEB data and obtain a historical session set;
[0008] Performing prediction on the historical conversation set using a web data analysis model to obtain a first conversation prediction result and a first user behavior safety prediction result;
[0009] performing at least one of determining a recommendation of the current WEB data and user monitoring according to the first session prediction result and the first user behavior safety prediction result;
[0010] The WEB data analysis model is obtained by the following steps:
[0011] Acquire a training data set, and extract a first session set from the training data set, where the first session set is a non-repeated first session;
[0012] Generate a first graph structure, embed the first conversation into the first graph structure, the first graph structure including a first node and a first edge, wherein the first node represents the first conversation, and the first edge represents a temporal relationship between the first conversations;
[0013] Adding a second node to the first graph structure, connecting the second node to all first nodes via a second edge to obtain a second graph structure, wherein the second node is used to transmit information of the unconnected first nodes;
[0014] Using a graph neural network to predict the embedded target on the second graph structure, to obtain a second session prediction result;
[0015] generating a third graph structure based on the second conversation prediction result, and performing prediction on the third graph structure using a graph autoencoder to obtain a second user behavior security prediction result;
[0016] The training parameters and the weight ratio of the training parameters of the WEB data analysis model are determined according to the second session prediction result and the second user behavior safety prediction result, and the WEB data analysis model is determined according to the training parameters and the weight ratio of the training parameters.
[0017] According to the user WEB data analysis method based on graph neural network, the method further includes:
[0018] Initializing the second graph structure, including initializing the first nodes and filling each first node with a different product;
[0019] The feature vectors of all first nodes are processed through a fully connected layer to obtain the initialization of the second nodes.
[0020] According to the user WEB data analysis method based on graph neural network, the method further includes:
[0021] Updating the first node and the second node using a message passing mechanism, including obtaining a first feature vector from an adjacent node using a message passing function for each first node;
[0022] performing aggregation processing on the second feature vector according to the first vector to obtain an updated node feature of the second node, wherein the aggregation processing includes one of mean aggregation and weighted aggregation;
[0023] The second node is updated according to the node feature of the updated second node to obtain a second feature vector, and the above updating steps are iteratively performed until the change value of the first feature vector and the second feature is less than a preset value.
[0024] According to the user WEB data analysis method based on graph neural network, the second graph structure is predicted by using graph neural network to embed the target, and the second session prediction result is obtained, including:
[0025] Using the updated second feature vector of the second node as a second graph structure representation, and selecting a plurality of the first nodes from the second graph structure according to user interests to generate a session representation;
[0026] Performing similarity calculation and user interest prediction on the session representations using a graph neural network, wherein the similarity calculation is performed by calculating the similarity between a first node and a second node in different sessions to determine the degree of association between the second node and the different session representations;
[0027] According to the similarity between the first node and the second node in the different sessions, the user's interest in the products in the different session representations is predicted to obtain a second prediction result.
[0028] According to the user WEB data analysis method based on graph neural network, a third graph structure is generated according to the second session prediction result, and a graph autoencoder is used to predict the third graph structure to obtain a second user behavior safety prediction result, including:
[0029] Generate a third graph structure based on the plurality of second prediction results according to a time sequence chain;
[0030] Obtaining neighbor node feature information of each node in the third graph structure through a graph autoencoder to generate a node representation of the third graph structure;
[0031] Reconstructing the node representation of the third graph structure using an autoencoder to obtain a reconstructed node representation of the third graph structure, and calculating a reconstruction error based on the node representation of the third graph structure and the reconstructed node representation of the third graph structure;
[0032] A second user behavior safety prediction result of the third graph structure is determined according to a preset error and the reconstruction error.
[0033] According to the user WEB data analysis method based on graph neural network, the autoencoder includes:
[0034] Obtain training data including normal purchases and security risk behaviors, convert the training data into low-dimensional vectors using a graph autoencoder, and use the graph autoencoder to predict the original graph and the reconstructed graph;
[0035] According to the difference metric determined between the original image and the reconstructed image, the loss function of the image autoencoder is determined according to the difference metric;
[0036] The autoencoder is back-propagated and updated according to the loss function, including the use of stochastic gradient descent to update the parameters of the autoencoder.
[0037] According to the user WEB data analysis method based on graph neural network, performing at least one of determining the recommendation of the current WEB data and user monitoring based on the first session prediction result and the first user behavior safety prediction result includes:
[0038] When the first session prediction result is a product of interest, the weight of the prediction of the product of interest to the user is increased; otherwise, the weight of the prediction of the product of interest to the user is reduced and the weight of the safety prediction of the user is increased.
[0039] Another aspect of an embodiment of the present invention provides a user WEB data analysis device based on a graph neural network, comprising:
[0040] The first module is used to obtain the user's historical WEB data and current WEB data, initialize the historical WEB data and obtain a historical session set;
[0041] The second module is configured to use a web data analysis model to predict the historical session set to obtain a first session prediction result and a first user behavior security prediction result;
[0042] A third module is configured to perform at least one of determining a recommendation of the current WEB data and user monitoring based on the first session prediction result and the first user behavior security prediction result;
[0043] The WEB data analysis model is obtained by training the fourth module, the fifth module, the sixth module, the seventh module, the eighth module and the ninth module:
[0044] The fourth module is used to obtain a training data set, and extract a first conversation set from the training data set, where the first conversation set is a non-repeated first conversation;
[0045] The fifth module is configured to generate a first graph structure, and embed the first conversation into the first graph structure, wherein the first graph structure includes a first node and a first edge, wherein the first node represents the first conversation, and the first edge represents a temporal relationship between the first conversations;
[0046] The sixth module is configured to add a second node to the first graph structure, wherein the second node is connected to all first nodes via a second edge to obtain a second graph structure, and the second node is configured to transmit information of the unconnected first nodes;
[0047] The seventh module is used to predict the embedded target using a graph neural network on the second graph structure to obtain a second session prediction result;
[0048] The eight modules are used to generate a third graph structure based on the second conversation prediction result, and use a graph autoencoder to predict the third graph structure to obtain a second user behavior safety prediction result;
[0049] The ninth module is used to determine the training parameters and the weight ratio of the training parameters of the WEB data analysis model based on the second session prediction result and the second user behavior safety prediction result, and determine the WEB data analysis model based on the training parameters and the weight ratio of the training parameters.
[0050] Another aspect of an embodiment of the present invention provides an electronic device, including a processor and a memory;
[0051] The memory is used to store programs;
[0052] The processor executes the program to implement the method described above.
[0053] Embodiments of the present invention further disclose a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device can read the computer instructions from the computer-readable storage medium and execute the computer instructions, causing the computer device to perform the method described above.
[0054] The beneficial effects of the present invention are: by adding an additional (second) node to the graph structure (graph data) corresponding to the user session, the recommendation effect of the user's products is improved; the product recommendation is based on the user's interests through the graph neural network, and the autoencoder is used to predict possible risky behaviors (such as artificial intelligence-based order brushing, invalid browsing and malicious comments, etc.), thereby improving the recommendation accuracy of the products that the user is interested in and improving the prediction accuracy of behavioral safety when purchasing products. BRIEF DESCRIPTION OF THE DRAWINGS
[0055] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the following description of the embodiments with reference to the accompanying drawings, in which:
[0056] Figure 1 This is a schematic diagram of the user WEB data analysis process based on graph neural network according to an embodiment of the present invention.
[0057] Figure 2 It is a schematic diagram of the WEB data analysis model initialization process according to an embodiment of the present invention.
[0058] Figure 3 It is a schematic diagram of the WEB data analysis model parameter update process according to an embodiment of the present invention.
[0059] Figure 4 2 is a flow chart of in-session commodity prediction according to an embodiment of the present invention.
[0060] Figure 5 1. It is a schematic diagram of the autoencoder security behavior prediction process of the WEB data analysis model according to an embodiment of the present invention.
[0061] Figure 6 2 is a schematic diagram of the autoencoder learning process according to an embodiment of the present invention.
[0062] Figure 7 This is a schematic diagram of a user WEB data analysis device based on a graph neural network according to an embodiment of the present invention. DETAILED DESCRIPTION
[0063] The following describes embodiments of the present invention in detail. Examples of the embodiments are shown in the accompanying drawings, where the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. In the subsequent description, suffixes such as "module," "component," or "unit" used to represent elements are used solely to facilitate the description of the present invention and have no specific meaning in themselves. Therefore, "module," "component," or "unit" may be used interchangeably. "First," "second," and the like are used solely to distinguish technical features and should not be construed as indicating or implying relative importance or implicitly indicating the number of the indicated technical features or implicitly indicating the order of the indicated technical features. In this subsequent description, the consecutive numbering of method steps is for ease of review and understanding. In conjunction with the overall technical solution of the present invention and the logical relationship between the various steps, adjusting the order of implementation of the steps does not affect the technical effects achieved by the technical solution of the present invention. The embodiments described below with reference to the accompanying drawings are exemplary and are intended only to explain the present invention and should not be construed as limiting the present invention.
[0064] refer to Figure 1 , Figure 1 This is a schematic diagram of a user WEB data analysis process based on a graph neural network according to an embodiment of the present invention. It includes but is not limited to steps S100 to S900:
[0065] S100 to S300 are the user WEB data analysis process, and S400 to S900 are the WEB data analysis model training diagrams. Specifically:
[0066] S100, obtaining the user's historical WEB data and current WEB data, initializing the historical WEB data and obtaining a historical session set.
[0067] S200 , predicting a historical session set using a WEB data analysis model to obtain a first session prediction result and a first user behavior safety prediction result.
[0068] S300 , performing at least one of determining a recommendation of current WEB data and user monitoring based on the first session prediction result and the first user behavior safety prediction result.
[0069] The WEB data analysis model is obtained through the following steps:
[0070] S400 , obtaining a training data set, and extracting a first conversation set from the training data set, where the first conversation set is a non-repeated first conversation.
[0071] S500: Generate a first graph structure, embed the first conversation into the first graph structure, the first graph structure includes a first node and a first edge, wherein the first node represents the first conversation, and the first edge represents the temporal relationship between the first conversations.
[0072] S600: Add a second node to the first graph structure. Connect the second node to all first nodes via a second edge to obtain a second graph structure. The second node is used to transmit information of unconnected first nodes.
[0073] In some embodiments, it can be understood that for any node A, it is not directly connected to node D (that is, there is no session association relationship), and a new session link is generated by connecting A-second node-D in time sequence through the second node to expand the original graph structure, so that a more accurate prediction effect can be obtained in subsequent session predictions.
[0074] In some embodiments, reference Figure 2 The schematic diagram of the WEB data analysis model initialization process shown includes but is not limited to steps S610 to S620:
[0075] S610, initializing the second graph structure, including initializing the first nodes, including filling each first node with a different product;
[0076] S620: Process the feature vectors of all first nodes through a fully connected layer to obtain initialization of the second node.
[0077] In some embodiments, reference Figure 3 The schematic diagram of the WEB data analysis model parameter update process shown includes but is not limited to steps S630 to S650:
[0078] S630, updating the first node and the second node using a message passing mechanism, including obtaining a first eigenvector from an adjacent node using a message passing function for each first node;
[0079] S640, performing aggregation processing on the second feature vector according to the first vector to obtain an updated node feature of the second node, wherein the aggregation processing includes one of mean aggregation and weighted aggregation;
[0080] S650, updating the node feature of the second node according to the updated second node to obtain a second feature vector, and iteratively performing the above updating steps until the change value of the first feature vector and the second feature vector is less than a preset value.
[0081] S700: Use a graph neural network to predict the embedded target of the second graph structure to obtain a second session prediction result.
[0082] In some embodiments, reference Figure 4 The flowchart of the product prediction process in the session shown in FIG. 1 includes but is not limited to steps S710 to S720:
[0083] S710, using the updated second feature vector of the second node as a second graph structure representation, and selecting multiple first nodes from the second graph structure based on user interests to generate a conversation representation;
[0084] S720, performing similarity calculation and user interest prediction on the conversation representations using a graph neural network, wherein the similarity calculation is performed by calculating the similarity between a first node and a second node in different conversations to determine the degree of association between the second node and the different conversation representations;
[0085] S730 : Predict the user's interest in the products in the different session representations based on the similarity between the first node and the second node in the different session, and obtain a second prediction result.
[0086] S800: Generate a third graph structure based on the second conversation prediction result, and use a graph autoencoder to predict the third graph structure to obtain a second user behavior safety prediction result.
[0087] In some embodiments, reference Figure 5 The schematic diagram of the autoencoder security behavior prediction process of the WEB data analysis model shown includes but is not limited to steps S810 to S840:
[0088] S810, generating a third graph structure based on a time series chain using the plurality of second prediction results;
[0089] S820, obtaining neighbor node feature information of each node in the third graph structure through a graph autoencoder, and generating a node representation of the third graph structure;
[0090] S830, reconstructing the node representation of the third graph structure using an autoencoder to obtain a reconstructed node representation of the third graph structure, and calculating a reconstruction error based on the node representation of the third graph structure and the reconstructed node representation of the third graph structure;
[0091] S840: Determine a second user behavior safety prediction result of the third graph structure according to the preset error and the reconstruction error.
[0092] In some embodiments, reference Figure 6 , Figure 6 Schematic diagram of the autoencoder learning process, which includes but is not limited to steps S850 to S870:
[0093] S850: Acquire training data including normal purchases of goods and security risk behaviors, convert the training data into low-dimensional vectors using a graph autoencoder, and predict the original graph and the reconstructed graph using the graph autoencoder;
[0094] S860, determining a difference metric between the original image and the reconstructed image, and determining a loss function of the image autoencoder according to the difference metric;
[0095] S870, performing backpropagation and updating processing on the autoencoder according to the loss function, including updating the parameters of the autoencoder using stochastic gradient descent.
[0096] S900, determining the training parameters and the weight ratio of the training parameters of the WEB data analysis model according to the second session prediction result and the second user behavior safety prediction result, and determining the WEB data analysis model according to the training parameters and the weight ratio of the training parameters.
[0097] Figure 7 1 is a diagram of a user web data analysis device based on a graph neural network according to an embodiment of the present invention. The device includes a first module 710, a second module 720, a third module 730, a fourth module 740, a fifth module 750, a sixth module 760, a seventh module 770, an eighth module 780, and a ninth module 790.
[0098] Among them, the first module is used to obtain the user's historical WEB data and current WEB data, initialize the historical WEB data and obtain the historical session set; the second module is used to predict the historical session set using the WEB data analysis model to obtain the first session prediction result and the first user behavior safety prediction result; the third module is used to perform at least one of the following processing: determining the recommendation of the current WEB data and user monitoring according to the first session prediction result and the first user behavior safety prediction result; wherein the WEB data analysis model is trained by the fourth module, the fifth module, the sixth module, the seventh module, the eighth module and the ninth module: the fourth module is used to obtain the training data set, extract the first session set from the training data set, and the first session set is a non-repeated first session; the fifth module is used to generate a first graph structure, embed the first session into the first graph structure, and the first The graph structure includes a first node and a first edge, wherein the first node represents a first session, and the first edge represents a temporal relationship between the first sessions; the sixth module is used to add a second node to the first graph structure, and the second node is connected to all the first nodes through the second edge to obtain a second graph structure, and the second node is used to transmit information of unconnected first nodes; the seventh module is used to use a graph neural network to predict the embedded target of the second graph structure to obtain a second session prediction result; the eighth module is used to generate a third graph structure based on the second session prediction result, and use a graph autoencoder to predict the third graph structure to obtain a second user behavior safety prediction result; the ninth module is used to determine the training parameters and the weight ratio of the training parameters of the WEB data analysis model based on the second session prediction result and the second user behavior safety prediction result, and determine the WEB data analysis model based on the training parameters and the weight ratio of the training parameters.
[0099] Exemplarily, with the cooperation of the first module in the device, the embodiment device can implement any of the aforementioned user WEB data analysis methods based on graph neural networks, namely, obtaining the user's historical WEB data and current WEB data, initializing the historical WEB data and obtaining a historical session set; using the WEB data analysis model to predict the historical session set to obtain a first session prediction result and a first user behavior safety prediction result; based on the first session prediction result and the first user behavior safety prediction result, performing at least one of the processes of determining the recommendation of the current WEB data and user monitoring; the WEB data analysis model is obtained by the following steps: obtaining a training data set, extracting a first session set from the training data set, the first session set being a non-repeated first session; generating a first graph structure, embedding the first session into the first A graph structure, the first graph structure includes a first node and a first edge, wherein the first node represents a first session, and the first edge represents a temporal relationship between the first sessions; a second node is added to the first graph structure, the second node is connected to all the first nodes through the second edge, to obtain a second graph structure, and the second node is used to transmit information of unconnected first nodes; a graph neural network is used to predict the embedded target of the second graph structure to obtain a second session prediction result; a third graph structure is generated according to the second session prediction result, and a graph autoencoder is used to predict the third graph structure to obtain a second user behavior safety prediction result; the training parameters of the WEB data analysis model and the weight ratio of the training parameters are determined according to the second session prediction result and the second user behavior safety prediction result, and the WEB data analysis model is determined according to the training parameters and the weight ratio of the training parameters. The beneficial effects of the present invention are: by adding an additional (second) node to the graph structure (graph data) corresponding to the user session, the recommendation effect of the user's products is improved; the product recommendation is based on the user's interests through the graph neural network, and the autoencoder is used to predict possible risky behaviors (such as artificial intelligence-based order brushing, invalid browsing and malicious comments, etc.), thereby improving the recommendation accuracy of the products that the user is interested in and improving the prediction accuracy of behavioral safety when purchasing products.
[0100] An embodiment of the present invention further provides an electronic device, the electronic device including a processor and a memory;
[0101] The memory stores a program;
[0102] The processor executes the program to perform the aforementioned user WEB data analysis method based on graph neural network; the electronic device has the function of carrying and running the software system for user WEB data analysis based on graph neural network provided by an embodiment of the present invention, such as a personal computer, minicomputer, main frame, workstation, network or distributed computing environment, separate or integrated computer platform, or communication with charged particle tools or other imaging devices, etc.
[0103] An embodiment of the present invention also provides a computer-readable storage medium, which stores a program, and the program is executed by a processor to implement the user WEB data analysis method based on graph neural network as described above.
[0104] In some optional embodiments, the function / operation mentioned in the block diagram may not occur in the order mentioned in the operation diagram. For example, depending on the function / operation involved, the two boxes shown in succession can actually be executed substantially simultaneously or the boxes can sometimes be executed in reverse order. In addition, the embodiment presented and described in the flow chart of the present invention is provided in an exemplary manner for the purpose of providing a more comprehensive understanding of the technology. The disclosed method is not limited to the operation and logic flow presented herein. Optional embodiments are contemplated in which the order of the various operations is changed and the sub-operations described as a part of a larger operation are performed independently.
[0105] Embodiments of the present invention further disclose a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device can read the computer instructions from the computer-readable storage medium and execute the computer instructions, causing the computer device to perform the aforementioned graph neural network-based user web data analysis method.
[0106] Furthermore, although the present invention is described in the context of functional modules, it should be understood that, unless otherwise indicated, one or more of the functions and / or features described may be integrated into a single physical device and / or software module, or one or more functions and / or features may be implemented in separate physical devices or software modules. It will also be understood that a detailed discussion of the actual implementation of each module is not necessary for understanding the present invention. More specifically, given the properties, functions, and internal relationships of the various functional modules in the devices disclosed herein, the actual implementation of the module will be understood within the ordinary skill of an engineer. Therefore, a person skilled in the art using ordinary skill will be able to implement the present invention set forth in the claims without undue experimentation. It will also be understood that the specific concepts disclosed are merely illustrative and are not intended to limit the scope of the present invention, which is determined by the full scope of the appended claims and their equivalents.
[0107] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.
[0108] The logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, an instruction execution system, apparatus, or device.
[0109] More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic devices), a portable computer disk cartridge (magnetic devices), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disc read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering, or processing in another suitable manner as necessary, and then stored in a computer memory.
[0110] It should be understood that various parts of the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application-specific integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.
[0111] Throughout this specification, reference to terms such as "one embodiment," "some embodiments," "examples," "specific examples," or "some examples" means that a specific feature, structure, material, or characteristic described in conjunction with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, schematic representations of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.
[0112] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to the embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the claims and their equivalents.
[0113] The above is a specific description of the preferred implementation of the present invention, but the present invention is not limited to the embodiments. Those skilled in the art can make various equivalent modifications or substitutions without violating the spirit of the present invention. These equivalent modifications or substitutions are all included in the scope defined by the claims of this application.
Claims
1. A user WEB data analysis method based on graph neural network, It is characterized in that include: Obtain the user's historical WEB data and current WEB data, initialize the historical WEB data and obtain a historical session set; Using a WEB data analysis model to predict the historical session set, to obtain a first session prediction result and a first user behavior safety prediction result; According to the first session prediction result and the first user behavior safety prediction result, performing at least one of a process of determining a recommendation of the current WEB data and a user monitoring; The WEB data analysis model is obtained by the following steps: Acquire a training data set, and extract a first session set from the training data set, where the first session set is a non-repeated first session; Generate a first graph structure, embed the first conversation into the first graph structure, the first graph structure includes a first node and a first edge, wherein the first node represents the first conversation, and the first edge represents a temporal relationship between the first conversations; Adding a second node to the first graph structure, wherein the second node is connected to all the first nodes through a second edge to obtain a second graph structure, wherein the second node is used to transmit information of the unconnected first nodes; Using a graph neural network to predict the embedded target on the second graph structure, to obtain a second session prediction result; Generate a third graph structure according to the second conversation prediction result, and use a graph autoencoder to predict the third graph structure to obtain a second user behavior security prediction result; The training parameters and the weight ratio of the training parameters of the WEB data analysis model are determined according to the second session prediction result and the second user behavior safety prediction result, and the WEB data analysis model is determined according to the training parameters and the weight ratio of the training parameters.
2. According to the user WEB data analysis method based on graph neural network according to claim 1, It is characterized in that The method further comprises: Initializing the second graph structure, including initializing the first nodes, including filling each first node with different commodities; The feature vectors of all first nodes are processed through a fully connected layer to obtain the initialization of the second nodes.
3. According to the user WEB data analysis method based on graph neural network according to claim 1, It is characterized in that The method further comprises: The first node and the second node are updated using a message passing mechanism, including obtaining a first feature vector from an adjacent node using a message passing function for each of the first nodes; Performing aggregation processing on the second feature vector according to the first vector to obtain an updated node feature of the second node, wherein the aggregation processing includes one of mean aggregation and weighted aggregation; The second node is updated according to the node feature of the updated second node to obtain a second feature vector, and the above updating step is iteratively performed until the change value of the first feature vector and the second feature is less than a preset value.
4. According to the user WEB data analysis method based on graph neural network according to claim 3, It is characterized in that The using a graph neural network to predict the embedded target on the second graph structure to obtain a second session prediction result includes: Using the updated second feature vector of the second node as a second graph structure representation, and selecting a plurality of the first nodes from the second graph structure according to user interests to generate a session representation; Performing similarity calculation and user interest prediction on the session representation using a graph neural network, wherein the similarity calculation determines the degree of association between the second node and the different session representations by calculating the similarity between the first node and the second node in different sessions; According to the similarity between the first node and the second node in different sessions, the user's interest in the commodities in the different session representations is predicted to obtain a second prediction result.
5. According to the user WEB data analysis method based on graph neural network according to claim 4, It is characterized in that The generating a third graph structure according to the second conversation prediction result, and predicting the third graph structure using a graph autoencoder to obtain a second user behavior safety prediction result, includes: Generate a third graph structure based on a time series chain using the plurality of the second prediction results; Obtaining neighbor node feature information of each node in the third graph structure through a graph autoencoder to generate a node representation of the third graph structure; Reconstructing the node representation of the third graph structure by using an autoencoder to obtain a reconstructed node representation of the third graph structure, and calculating a reconstruction error according to the node representation of the third graph structure and the reconstructed node representation of the third graph structure; According to the preset error and the reconstruction error, a second user behavior safety prediction result of the third graph structure is determined.
6. According to the user WEB data analysis method based on graph neural network according to claim 5, It is characterized in that The autoencoder comprises: Obtain training data including normal purchase of goods and security risk behaviors, convert the training data into low-dimensional vectors using a graph autoencoder, and predict the original graph and reconstructed graph through the graph autoencoder; According to the determined difference metric between the original image and the reconstructed image, the loss function of the image autoencoder is determined according to the difference metric; The autoencoder is back-propagated and updated according to the loss function, including updating the parameters of the autoencoder using stochastic gradient descent.
7. According to the user WEB data analysis method based on graph neural network according to claim 1, It is characterized in that The performing at least one of determining the recommendation of the current WEB data and user monitoring according to the first session prediction result and the first user behavior security prediction result includes: When the first session prediction result is a product of interest, the weight of the prediction of the product of interest to the user is increased; otherwise, the weight of the prediction of the product of interest to the user is reduced and the weight of the safety prediction of the user is increased.
8. A user WEB data analysis device based on graph neural network, It is characterized in that include: The first module is used to obtain the user's historical WEB data and current WEB data, initialize the historical WEB data and obtain a historical session set; The second module is used to predict the historical session set using a WEB data analysis model to obtain a first session prediction result and a first user behavior safety prediction result; A third module is used to perform at least one of determining the recommendation of the current WEB data and user monitoring according to the first session prediction result and the first user behavior security prediction result; The WEB data analysis model is obtained by training the fourth module, the fifth module, the sixth module, the seventh module, the eighth module and the ninth module: The fourth module is used to obtain a training data set, and extract a first conversation set from the training data set, where the first conversation set is a non-repeating first conversation; The fifth module is used to generate a first graph structure, embed the first conversation into the first graph structure, the first graph structure includes a first node and a first edge, wherein the first node represents the first conversation, and the first edge represents the temporal relationship between the first conversations; The sixth module is used to add a second node to the first graph structure, the second node is connected to all the first nodes through a second edge to obtain a second graph structure, and the second node is used to transmit information of the unconnected first nodes; The seventh module is used to predict the embedded target using a graph neural network on the second graph structure to obtain a second session prediction result; The eight modules are used to generate a third graph structure according to the second session prediction result, and predict the third graph structure using a graph autoencoder to obtain a second user behavior safety prediction result; The ninth module is used to determine the training parameters and the weight ratio of the training parameters of the WEB data analysis model according to the second session prediction result and the second user behavior safety prediction result, and determine the WEB data analysis model according to the training parameters and the weight ratio of the training parameters.
9. An electronic device, It is characterized in that including a processor and a memory; The memory is used to store programs; The processor executes the program to implement the user WEB data analysis method based on graph neural network as described in any one of claims 1-7.
10. A computer-readable storage medium, It is characterized in that The storage medium stores a program, and the program is executed by a processor to implement the user WEB data analysis method based on graph neural network as described in any one of claims 1 to 7.