Multi-behavior sequence recommendation method and system based on graph neural network

Abstract

The invention discloses a multi-behavior sequence recommendation method and system based on a graph neural network, and relates to the technical field of recommendation systems.The method comprises the steps that a global article graph is constructed through multi-behavior sequence data, and a user interaction graph is obtained based on interaction of a user in the global article graph; constructing a user node feature matrix and an article node feature matrix in the user interaction graph, performing feature learning on the user interaction graph by using a graph neural network, and updating representation of nodes; on the basis of the updated user interaction graph, generating a personalized recommendation list by utilizing a reinforcement learning algorithm and combining feedback of the user; through the graph neural network and the reinforcement learning algorithm, the problems of cold start and reduction of recommendation accuracy in a traditional recommendation system can be effectively solved, and a recommendation result with more pertinence and timeliness is provided for a user, so that the user experience and satisfaction are enhanced.

Description

Technical Field

[0001] The present invention relates to the technical field of recommendation systems, and in particular to a multi-behavior sequence recommendation method and system based on graph neural network. Background Art

[0002] In recent years, with the development of the Internet and big data technology, recommendation systems have become an indispensable component of digital platforms. Traditional recommendation algorithms mainly rely on users' historical behavior data, such as click, browse and purchase records, and usually use collaborative filtering or content-based methods. However, these methods have certain limitations in dealing with diverse user behaviors and complex relationships. In addition, the shortcomings of the existing technology are that traditional recommendation systems usually ignore the multi-behavior sequence characteristics of users and the relationship network between items. These features are crucial for accurately understanding user preferences. Specifically, existing methods often fail to effectively integrate multiple user behaviors (such as clicks, purchases, ratings, etc.) to construct a more comprehensive user portrait. The user interaction graph is not dynamic enough and cannot reflect changes in user preferences in real time, resulting in limited timeliness and accuracy of recommendation results. Summary of the invention

[0003] In view of the above existing problems, the present invention is proposed.

[0004] Therefore, the present invention provides a multi-behavior sequence recommendation method and system based on graph neural network to solve the problem of user behavior diversity and dynamic interactive changes affecting recommendation accuracy.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0006] In a first aspect, an embodiment of the present invention provides a multi-behavior sequence recommendation method based on a graph neural network, which includes constructing a global item graph using multi-behavior sequence data, and obtaining a user interaction graph based on user interactions in the global item graph;

[0007] In the user interaction graph, construct the user node feature matrix and the item node feature matrix, and use the graph neural network to learn the features of the user interaction graph and update the node representation;

[0008] Based on the updated user interaction graph, a reinforcement learning algorithm is used to generate a personalized recommendation list in combination with user feedback.

[0009] As a preferred solution of the multi-behavior sequence recommendation method based on graph neural network described in the present invention, the multi-behavior sequence data includes the number of clicks, the number of purchases, the rating of the item, the item category, the user ID and the item ID.

[0010] As a preferred solution of the multi-behavior sequence recommendation method based on graph neural network described in the present invention, wherein: using multi-behavior sequence data to construct a global item graph includes the following steps:

[0011] Extract all item IDs from multi-behavior sequence data as item nodes;

[0012] Define the relationship between items, extract item categories from the user's multi-behavior sequence data based on the defined relationship, and establish edges between item nodes;

[0013] Using the constructed item nodes and edges, the graph database is used to create the initial structure of the global item graph. The number of co-purchases is used as the edge weight to indicate how often two items are purchased together. The similarity of click counts is used to calculate the edge weight to indicate the user's attention to the item.

[0014] Based on changes in user behavior, a dynamic update mechanism is designed to adjust the global item map in real time, use visualization tools to display the relationship between items, and output the completed global item map.

[0015] As a preferred solution of the multi-behavior sequence recommendation method based on graph neural network described in the present invention, wherein: based on the user's interaction in the global item graph, obtaining the user interaction graph includes the following steps:

[0016] Extract all user IDs from multi-behavior sequence data as user nodes, and item IDs as item nodes. Create edges based on the interaction between users and items, and use the graph database to build a user interaction graph.

[0017] Using the number of new interactions between users and items, we design a dynamic and adaptive graph update mechanism to update edge weights, expressed as,

[0018] w′=w+α×Δ;

[0019] Among them, w′ represents the updated edge weight, w represents the initial weight, α represents the adjustment factor, and Δ represents the number of new interactions between users and items;

[0020] Based on the update of edge weights, the updated user interaction graph is obtained.

[0021] As a preferred solution of the multi-behavior sequence recommendation method based on graph neural network described in the present invention, in which: in the user interaction graph, constructing the user node feature matrix and the item node feature matrix includes the following steps:

[0022] Based on the user interaction graph, count the number of historical interactions;

[0023] Count the item categories that users interact with the most, and represent them with one-hot encoding to get the preferred item categories;

[0024] Collect the timestamp of the user's most recent interaction to get the most recent interaction time;

[0025] Integrate the historical number of interactions, preferred item categories, and the most recent interaction time to obtain the user node feature matrix;

[0026] Based on the user interaction graph, count the number of times items are interacted with;

[0027] Get the item category in the global item graph, use one-hot encoding to represent it, and get the item attributes;

[0028] Count the average ratings of all users for the same item;

[0029] The number of interactions, item attributes and average ratings are integrated to obtain the item node feature matrix.

[0030] As a preferred solution of the multi-behavior sequence recommendation method based on graph neural network described in the present invention, wherein: based on the user node feature matrix and the item node feature matrix, the graph neural network is used to perform feature learning on the user interaction graph and update the node representation, including the following steps:

[0031] Define an adjacency matrix to represent the user interaction graph and perform normalization;

[0032] In the graph convolutional network, the update formula of the user and item node feature matrix is ​​expressed as,

[0033]

[0034] Among them, H (l+1) represents the feature matrix of user and item nodes at the l+1th layer, H (l) represents the feature matrix of user and item nodes at the lth layer, W (l) represents the learned weight matrix, σ represents the activation function, represents the normalized adjacency matrix;

[0035] The initial feature matrix is ​​formed by the feature matrices of user and item nodes;

[0036] Through multi-layer graph convolution operations, the feature matrices of users and items are aggregated, and new representations of each user node and item node are output.

[0037] As a preferred solution of the multi-behavior sequence recommendation method based on graph neural network described in the present invention, wherein: based on the updated user interaction graph, using reinforcement learning algorithm, and combining user feedback, a personalized recommendation list is generated, including the following steps:

[0038] Based on the updated user interaction graph, the updated feature vectors of the user and item nodes are obtained. The updated feature vectors of the user nodes and the user behavior at the current time are used to form the state space. The updated feature vectors of the item nodes are used to form a recommendation list to form the action space. Based on user feedback, a reward mechanism is set.

[0039] Using DQN, calculate the Q value of taking an action in a state and update the Q value using the Bellman equation;

[0040] Use the ε-greedy strategy to balance exploration and exploitation to determine action selection;

[0041] Based on the learned Q value, the probability distribution of recommended items is output, and the recommendation list with the largest Q value is selected. When generating the recommendation list, the user's real-time behavior data is considered, the recommended content is dynamically adjusted, and a personalized recommendation list containing the ID and name of the recommended item is generated.

[0042] In a second aspect, the present invention provides a multi-behavior sequence recommendation system based on a graph neural network, including: a user interaction graph construction module, which is responsible for constructing a global item graph using multi-behavior sequence data, and obtaining a user interaction graph based on user interactions in the global item graph;

[0043] The user interaction graph feature learning module is responsible for constructing the user node feature matrix and the item node feature matrix in the user interaction graph, and using the graph neural network to learn the features of the user interaction graph and update the node representation;

[0044] The personalized recommendation list generation module is responsible for generating a personalized recommendation list based on the updated user interaction graph, using the reinforcement learning algorithm and combining user feedback.

[0045] In a third aspect, an embodiment of the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, any step of the multi-behavior sequence recommendation method based on graph neural network as described in the first aspect of the present invention is implemented.

[0046] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, any step of the multi-behavior sequence recommendation method based on graph neural network as described in the first aspect of the present invention is implemented.

[0047] The beneficial effects of the present invention are: using graph neural networks to perform feature learning on user interaction graphs, so that node representations can better reflect the relationship between users and items. This method not only takes into account the historical behavior of users, but also introduces the relationship between items, thereby capturing deeper interaction patterns in the feature learning process. Through the reinforcement learning algorithm, combined with real-time feedback from users, the system can dynamically adjust the recommendation strategy to ensure that the recommended content better meets the real-time needs of users. Using the DQN algorithm, the system realizes intelligent decision-making in state space and action space, thereby improving the accuracy and personalization of recommendations. Ultimately, the present invention can effectively solve the cold start problem and the problem of decreased recommendation accuracy in traditional recommendation systems, and provide users with more targeted and timely recommendation results, thereby enhancing user experience and satisfaction. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0049] Figure 1 This is a flowchart of the multi-behavior sequence recommendation method based on graph neural network in Example 1.

[0050] Figure 2 This is a flowchart of the multi-behavior sequence recommendation system based on graph neural network in Example 1. DETAILED DESCRIPTION

[0051] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0052] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0053] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0054] Example 1, reference Figure 1 and Figure 2, which is the first embodiment of the present invention, and provides a multi-behavior sequence recommendation method based on graph neural network, comprising the following steps:

[0055] S1. Using multi-behavior sequence data, construct a global item graph, and obtain a user interaction graph based on the user's interaction in the global item graph, including the following steps:

[0056] S1.1. Multi-behavior sequence data includes the number of clicks, the number of purchases, the ratings of items, the item categories, the user IDs and the item IDs.

[0057] It should be noted that by integrating multiple behavioral sequence data such as the number of user clicks, purchases, ratings, item categories, user IDs, and item IDs, a comprehensive capture of user behavior is achieved. This data integration provides a richer feature basis for the subsequent recommendation system, enabling the system to more accurately analyze user preferences and ultimately improve the accuracy and relevance of recommendations.

[0058] Preprocess the multi-behavior sequence data, including deleting duplicate interaction records, unifying the timestamp format, and ensuring the time order of all behavior records, and extracting features from the original multi-behavior sequence data, including user ID, item ID, behavior type, and timestamp;

[0059] Extract all item IDs from multi-behavior sequence data as item nodes. Nodes represent all recommended items, such as movies, books, and commodities. Ensure the integrity of item nodes and avoid missing newly added items.

[0060] Define the relationship between items, including defining co-purchase relationships based on items purchased by users at the same time, defining co-click relationships based on items clicked by users at similar times, defining similar attribute relationships based on item categories, and defining rating relationships based on the similarity of user ratings of items. Based on the defined relationships, extract item categories from the user's multi-behavior sequence data and establish edges between item nodes;

[0061] Using the constructed item nodes and edges, use a graph database (such as NetworkX) to create the initial structure of the global item graph. In the graph, each item node is connected to other related item nodes to form a network. Each edge is attached with a weight to indicate the strength of the relationship between items. The number of co-purchases is used as the edge weight to indicate how often two items are purchased together. The similarity of the number of clicks is used to calculate the edge weight to indicate the degree of user attention to the item.

[0062] Based on changes in user behavior, a dynamic update mechanism is designed to adjust the global item graph in real time. When users interact with new items, the related co-purchase and co-click relationships are updated. The relationship weights between items are recalculated regularly to reflect the latest user behavior data. For items that have not been interacted with for a long time, you can consider reducing their weight in the graph to avoid affecting the recommendation effect. Use visualization tools to display the relationship between items and output the constructed global item graph.

[0063] It should be noted that the global item graph is constructed using multi-behavior sequence data, and the structure of item nodes and edges is established by extracting item IDs and defining the relationship between items. This process realizes the visualization and dynamic monitoring of the relationship between items, allowing the recommendation system to better understand the mutual influence between items, so as to take into account the relevance of items when recommending, thereby achieving the effect of increasing user click-through rate and purchase rate.

[0064] S1.2. Extract all user IDs from the multi-behavior sequence data as user nodes and item IDs as item nodes. Create edges based on the interaction behaviors between users and items, and use a graph database (such as Neo4j, NeoTech) to build a user interaction graph.

[0065] Monitor user interaction behaviors in real time, such as new behaviors of users on the platform (such as new clicks, new purchases). When new interactions occur, trigger the graph update process. When users frequently interact with a certain type of item, increase the weight of the item. Use the number of new interactions between users and items to design a dynamic and adaptive graph update mechanism to update the edge weights, expressed as,

[0066] w′=w+α×Δ;

[0067] Among them, w′ represents the updated edge weight, w represents the initial weight, α represents the adjustment factor (which can be set to 0.1), and Δ represents the number of new interactions between users and items;

[0068] If the number of interactions between a user and a certain type of item exceeds the set threshold (such as 5 times), a new edge is automatically added to indicate a strong preference relationship. Conversely, if the number of interactions is lower than the threshold (such as 1 time), the edge will be considered for deletion. After all edge weight adjustments and edge addition and deletion operations, the structure of the graph is updated to ensure that the graph always reflects the latest user preferences. Based on the update of edge weights, the updated user interaction graph is obtained.

[0069] It should be noted that based on the global item graph, the user ID and its interaction behavior with items are extracted to construct the user interaction graph. Through the dynamic and adaptive graph update mechanism, the latest interaction behavior of users is reflected in a timely manner. This mechanism ensures the timeliness of the user interaction graph and can quickly respond to the changing preferences of users, thereby enhancing the user experience and improving the real-time and personalization of recommendations.

[0070] S2. In the user interaction graph, construct the user node feature matrix and the item node feature matrix, and use the graph neural network to learn the features of the user interaction graph and update the node representation, including the following steps:

[0071] S2.1. Based on the user interaction graph, the historical interaction times are counted and expressed as:

[0072] I(u i )=∑ j I(u i ,p j );

[0073] Among them, I(u i ) represents user u i The total number of interactions with all items, u i represents the i-th user node, p j represents the jth item node, I(u i ,p j ) represents user u i With item p j The number of interactions between them, i represents the user node index, and j represents the item node index;

[0074] Count the item categories that users interact with the most, and represent them with one-hot encoding to get the preferred item categories;

[0075] Collect the timestamp of the user's most recent interaction, normalize it to a value between 0 and 1 to represent the user's activity, and get the most recent interaction time;

[0076] Integrate the historical number of interactions, preferred item categories, and the most recent interaction time to obtain the user node feature matrix;

[0077] Based on the user interaction graph, the number of item interactions is counted, expressed as,

[0078] p(p j )=∑ i I(u i ,p j );

[0079] Among them, p(p j ) represents item p j Number of interactions;

[0080] Get the item category in the global item graph, use one-hot encoding to represent it, and get the item attributes;

[0081] Count the average rating of all users for the same item, expressed as,

[0082]

[0083] Among them, A(p j ) represents the average rating of all users for the same item, r(u i ,p j ) represents user u i For item p j Rating, R(p j ) represents the j The number of users who have made ratings;

[0084] The number of interactions, item attributes and average ratings are integrated to obtain the item node feature matrix.

[0085] It should be noted that based on the user interaction graph, the feature matrix of users and items is constructed, and a comprehensive feature description of users and items is achieved by counting historical interaction times, preferred item categories, and recent interaction time. The construction of this feature matrix provides a basis for subsequent graph neural network feature learning, ensuring that the system can fully utilize the user's historical behavior information during the learning process, thereby improving the accuracy of recommendations.

[0086] S2.2, define the adjacency matrix to represent the user interaction graph and perform normalization, where the interaction between the user and the item is represented by 1, and the node without interaction is represented by 0;

[0087] In the graph convolutional network, the update formula of the user and item node feature matrix is ​​expressed as,

[0088]

[0089] Among them, H (l+1) represents the feature matrix of the user and item nodes at the l+1th layer, that is, the updated user and item node representations after output, H (l+1) The dimension is N×F (l+1) , N represents the number of nodes, F (l+1) represents the feature dimension of the l+1th layer, H (l) Represents the feature matrix of user and item nodes at the lth layer, where each row corresponds to a feature vector of a user and item node, and the dimension of the feature matrix is ​​N×F (l) , F (l) is the feature dimension of this layer, W (l) Represents the learned weight matrix, dimension F (l) ×F (l +1) , used to map the features of the current layer to the feature space of the next layer, σ represents the activation function ReLU (Rectified Linear Unit), expressed as σ(x) = max(0,x), Represents the normalized adjacency matrix, which is used to represent the connection relationship between nodes and ensure that the weights of neighboring nodes are considered when aggregating features;

[0090] Set the number of layers, usually 2 or 3, to balance the computational complexity and the expressiveness of the graph convolutional network. The more layers there are, the more complex feature relationships the model can capture, but the computational cost and the risk of overfitting will also increase.

[0091] The initial feature matrix is ​​composed of the feature matrices of user and item nodes, expressed as,

[0092]

[0093] Among them, H (0) represents the initial feature matrix, U represents the user node feature matrix, and P represents the item node feature matrix;

[0094] Through multi-layer graph convolution operations, the feature matrices of users and items are aggregated, and new representations of each user node and item node are output, expressed as,

[0095]

[0096]

[0097] in, represents the updated feature vector of the i-th user node, represents the updated feature vector of the jth item node, m represents the number of user nodes, H (L) Represents the node feature matrix of the Lth layer output of the graph convolutional network, which contains the feature representation of all nodes after L layers of graph convolution operations. [i] represents the feature matrix H (L) Extract the i-th row from the feature matrix, which represents the i-th user node. Since the index of the user node in the feature matrix is ​​continuous, it can be directly accessed through [i]. [m+j] represents the index of the user node in the feature matrix H. (L) Extract the m+jth row, which represents the jth item node.

[0098] It should be noted that the representation of user and item nodes is updated by learning features of the user interaction graph through graph neural networks. In this process, the normalization of the adjacency matrix and the graph convolution operation enable the node features to effectively aggregate the information of the surrounding neighbors. This feature learning process enables the model to capture richer user preferences and item characteristics, thereby improving the relevance of recommendations and user satisfaction.

[0099] S3. Based on the updated user interaction graph, using the reinforcement learning algorithm and combining user feedback, a personalized recommendation list is generated, including the following steps:

[0100] Based on the updated user interaction graph, the updated feature vectors of the user and item nodes are obtained. The updated feature vectors of the user nodes and the user behavior at the current time are used to form the state space. The updated feature vectors of the item nodes are used to form a recommendation list to form the action space. Based on user feedback, a reward mechanism is set. If the user clicks or buys the recommended item, the reward is 1, otherwise it is 0. The discount factor is used to calculate the discounted value of future rewards to enhance the learning of long-term behavior.

[0101] Using DQN (Deep Q-Network), calculate the Q value of taking action in the state, and use the Bellman equation to update the Q value, expressed as,

[0102]

[0103] Among them, Q(s t ,a t ) means in state s t Take action a t The current Q value, α represents the learning rate, which is a hyperparameter between 0 and 1, controlling the influence of new information on the current Q value, s t represents the state space, a t represents the action space, r t represents the immediate reward obtained by the user after interacting with the recommended item at time step t, γ represents the discount factor, and s t+1 represents the next state after time step t, Indicates that in state s t+1 The maximum value of the Q value of all possible actions a;

[0104] Use the ε-greedy strategy to balance exploration and exploitation to determine action selection;

[0105] Based on the learned Q-value, the probability distribution of recommended items is output, and the recommendation list with the largest Q-value is selected. When generating the recommendation list, the user's real-time behavior data (such as recent clicks, browsing history, etc.) is considered, and the recommended content is dynamically adjusted to generate a personalized recommendation list containing the ID and name of the recommended item.

[0106] It should be noted that based on the updated user interaction graph and combined with user feedback, a personalized recommendation list is generated using a reinforcement learning algorithm. Through the application of DQN and the Bellman equation, the system is able to strike a balance between exploration and exploitation and adjust the recommendation strategy in real time. This process achieves dynamic adaptation to user preferences, ensuring that the recommended content can accurately match user needs, thereby increasing user click and purchase rates and ultimately enhancing business conversion effects.

[0107] This embodiment also provides a multi-behavior sequence recommendation system based on graph neural network, including: a user interaction graph construction module, responsible for using multi-behavior sequence data to construct a global item graph, and obtaining a user interaction graph based on user interactions in the global item graph;

[0108] The user interaction graph feature learning module is responsible for constructing the user node feature matrix and the item node feature matrix in the user interaction graph, and using the graph neural network to learn the features of the user interaction graph and update the node representation;

[0109] The personalized recommendation list generation module is responsible for generating a personalized recommendation list based on the updated user interaction graph, using the reinforcement learning algorithm and combining user feedback.

[0110] The user interaction graph construction module is responsible for building a global item graph using multi-behavior sequence data and obtaining a user interaction graph based on the user's interaction in the global item graph;

[0111] The user interaction graph feature learning module is responsible for constructing the user node feature matrix and the item node feature matrix in the user interaction graph, and using the graph neural network to learn the features of the user interaction graph and update the node representation;

[0112] The personalized recommendation list generation module is responsible for generating a personalized recommendation list based on the updated user interaction graph, using the reinforcement learning algorithm and combining user feedback.

[0113] This embodiment also provides a computer device, which is suitable for the case of a multi-behavior sequence recommendation method based on a graph neural network, including: a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the multi-behavior sequence recommendation method based on a graph neural network as proposed in the above embodiment.

[0114] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0115] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the method for implementing a multi-behavior sequence recommendation based on a graph neural network as proposed in the above embodiment is implemented; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as a static random access memory (SRAM), an electrically erasable programmable read-only memory (EEPROM), an erasable programmable read-only memory (EPROM), a programmable read-only memory (PROM), a read-only memory (ROM), a magnetic memory, a flash memory, a disk or an optical disk.

[0116] In summary, the present invention achieves this by: using graph neural networks to perform feature learning on user interaction graphs, so that node representations can better reflect the relationship between users and items. This method not only takes into account the historical behavior of users, but also introduces the relationship between items, thereby capturing deeper interaction patterns in the feature learning process. Through the reinforcement learning algorithm, combined with real-time feedback from users, the system can dynamically adjust the recommendation strategy to ensure that the recommended content better meets the real-time needs of users. Using the DQN algorithm, the system implements intelligent decision-making in state space and action space, thereby improving the accuracy and personalization of recommendations. Ultimately, the present invention can effectively solve the cold start problem and the problem of decreased recommendation accuracy in traditional recommendation systems, and provide users with more targeted and timely recommendation results, thereby enhancing user experience and satisfaction.

Claims

1. A multi-behavior sequence recommendation method based on graph neural network, characterized by: include, Utilize multi-behavior sequence data to build a global item graph, and obtain a user interaction graph based on the user's interaction in the global item graph; In the user interaction graph, construct the user node feature matrix and the item node feature matrix, and use the graph neural network to learn the features of the user interaction graph and update the node representation; Based on the updated user interaction graph, a reinforcement learning algorithm is used to generate a personalized recommendation list in combination with user feedback.

2. The multi-behavior sequence recommendation method based on graph neural network according to claim 1, characterized in that: The multi-behavior sequence data includes the number of clicks, the number of purchases, the rating of the item, the item category, the user ID and the item ID.

3. The multi-behavior sequence recommendation method based on graph neural network as claimed in claim 2, characterized in that: Using multi-behavior sequence data, constructing a global item graph includes the following steps: Extract all item IDs from multi-behavior sequence data as item nodes; Define the relationship between items, extract item categories from the user's multi-behavior sequence data based on the defined relationship, and establish edges between item nodes; Using the constructed item nodes and edges, the graph database is used to create the initial structure of the global item graph. The number of co-purchases is used as the edge weight to indicate how often two items are purchased together. The similarity of click counts is used to calculate the edge weight to indicate the user's attention to the item. Based on changes in user behavior, a dynamic update mechanism is designed to adjust the global item map in real time, use visualization tools to display the relationship between items, and output the completed global item map.

4. The multi-behavior sequence recommendation method based on graph neural network as claimed in claim 3, characterized in that: Based on the user's interaction in the global item graph, a user interaction graph is obtained, including the following steps: Extract all user IDs from multi-behavior sequence data as user nodes, and item IDs as item nodes. Create edges based on the interaction between users and items, and use the graph database to build a user interaction graph. Using the number of new interactions between users and items, we design a dynamic and adaptive graph update mechanism to update edge weights, expressed as, w′=w+α×Δ; Among them, w′ represents the updated edge weight, w represents the initial weight, α represents the adjustment factor, and Δ represents the number of new interactions between users and items; Based on the update of edge weights, the updated user interaction graph is obtained.

5. The multi-behavior sequence recommendation method based on graph neural network according to claim 4, characterized in that: In the user interaction graph, constructing the user node feature matrix and the item node feature matrix includes the following steps: Based on the user interaction graph, count the number of historical interactions; Count the item categories that users interact with the most, and represent them with one-hot encoding to get the preferred item categories; Collect the timestamp of the user's most recent interaction to get the most recent interaction time; Integrate the historical number of interactions, preferred item categories, and the most recent interaction time to obtain the user node feature matrix; Based on the user interaction graph, count the number of times items are interacted with; Get the item category in the global item graph, use one-hot encoding to represent it, and get the item attributes; Count the average ratings of all users for the same item; The number of interactions, item attributes and average ratings are integrated to obtain the item node feature matrix.

6. The multi-behavior sequence recommendation method based on graph neural network according to claim 5, characterized in that: Based on the user node feature matrix and the item node feature matrix, the graph neural network is used to learn the features of the user interaction graph and update the node representation, including the following steps: Define an adjacency matrix to represent the user interaction graph and perform normalization; In the graph convolutional network, the update formula of the user and item node feature matrix is ​​expressed as, Among them, H (l+1) represents the feature matrix of user and item nodes at the l+1th layer, H (l) represents the feature matrix of user and item nodes at the lth layer, W (l) represents the learned weight matrix, σ represents the activation function, represents the normalized adjacency matrix; The initial feature matrix is ​​formed by the feature matrices of user and item nodes; Through multi-layer graph convolution operations, the feature matrices of users and items are aggregated, and new representations of each user node and item node are output.

7. The multi-behavior sequence recommendation method based on graph neural network according to claim 6, characterized in that: Based on the updated user interaction graph, a reinforcement learning algorithm is used to generate a personalized recommendation list in combination with user feedback, including the following steps: Based on the updated user interaction graph, the updated feature vectors of the user and item nodes are obtained. The updated feature vectors of the user nodes and the user behavior at the current time are used to form the state space. The updated feature vectors of the item nodes are used to form a recommendation list to form the action space. Based on user feedback, a reward mechanism is set. Using DQN, calculate the Q value of taking an action in a state and update the Q value using the Bellman equation; Use the ε-greedy strategy to balance exploration and exploitation to determine action selection; Based on the learned Q value, the probability distribution of recommended items is output, and the recommendation list with the largest Q value is selected. When generating the recommendation list, the user's real-time behavior data is considered, the recommended content is dynamically adjusted, and a personalized recommendation list containing the ID and name of the recommended item is generated.

8. A multi-behavior sequence recommendation system based on a graph neural network, based on the multi-behavior sequence recommendation method based on a graph neural network according to any one of claims 1 to 7, characterized in that: include, The user interaction graph construction module is responsible for building a global item graph using multi-behavior sequence data and obtaining a user interaction graph based on the user's interaction in the global item graph; The user interaction graph feature learning module is responsible for constructing the user node feature matrix and the item node feature matrix in the user interaction graph, and using the graph neural network to learn the features of the user interaction graph and update the node representation; The personalized recommendation list generation module is responsible for generating a personalized recommendation list based on the updated user interaction graph, using the reinforcement learning algorithm and combining user feedback.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the multi-behavior sequence recommendation method based on graph neural network described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the multi-behavior sequence recommendation method based on graph neural network described in any one of claims 1 to 7 are implemented.

Images