Intelligent recommendation method and device for member behavior analysis, equipment and storage medium
By integrating multi-source heterogeneous data and using graph neural networks and multi-task learning optimization models, the problem of insufficient system cold start and data real-time processing capabilities is solved, efficient and accurate personalized recommendations are achieved, and the timeliness and accuracy of member behavior analysis is improved.
Patent Information
- Application Number
- CN202510320577.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-18
- Publication Date
- 2025-08-01
AI Technical Summary
In the face of the complex situation of multi-source heterogeneous data, the existing technology is difficult to solve the system cold start problem, and the data standardization and real-time processing capabilities are insufficient, resulting in poor timeliness and accuracy of behavioral analysis results, making it difficult to achieve efficient and accurate personalized recommendations.
Through the ETL process, data cleaning and enhancement are carried out, membership relationships and behavior patterns are captured using graph neural network, multi-task learning and self-supervised learning optimized behavior analysis models, time series prediction models are constructed, and personalized recommendations are generated.
Real-time analysis and personalized recommendation of member behavior are realized, work efficiency and member satisfaction are improved, and the problems of insufficient data fusion and real-time in the existing technology are solved.
Smart Images
Figure CN120407546A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of data analysis, and particularly relates to an intelligent recommendation method, device, equipment and storage medium for member behavior analysis. Background Art
[0002] With the development of big data technology, more and more fields begin to use it for user portrait analysis or behavior analysis. The current user portrait and behavior analysis technologies mainly rely on traditional data mining and machine learning methods. Usually, after collecting the basic information of users, clustering analysis, classification algorithms and other technologies are used to construct user feature models. Especially in organizational management and member education work, personalized recommendation for the behavior and interests of individual members is an important means to improve work efficiency and member satisfaction.
[0003] However, the existing technologies have significant deficiencies when facing multi-source heterogeneous data, such as groups that need to carry out a large number of collaborative activities but the actual situations of individual members are different. The existing technologies are difficult to solve the system cold start problem in the initial stage of the system, that is, it is difficult to make accurate recommendations when there is insufficient user data in the initial stage of the system, and the data standardization and real-time processing capabilities are weak, resulting in poor timeliness and accuracy of the results of behavior analysis and recommended content, and it is difficult to achieve efficient personalized recommendation. A single feature extraction method cannot fully mine the deep information in the data, which limits the performance of the recommendation system. The existing technologies have obvious defects in data fusion, model complexity and real-time performance, and it is difficult to meet the requirements of efficient and accurate personalized recommendation.
[0004] Therefore, there is a need for a technical means that can effectively perform behavior analysis and intelligent recommendation when facing complex data. Summary of the Invention
[0005] The main object of the present invention is to provide an intelligent recommendation method, device, equipment and storage medium for member behavior analysis, aiming to solve the problem that the existing technical means have obvious defects in data fusion, model complexity and real-time performance, and it is difficult to meet the requirements of efficient and accurate personalized recommendation.
[0006] To achieve the above object, the present invention proposes an intelligent recommendation method, device, equipment and storage medium for member behavior analysis. The method includes the following steps: Collect member behavior data, integrate organizational activity records, integrate multi-source heterogeneous data of members' data on the online learning platform and members' platform social interactions, and then perform data integration through the ETL process to obtain initial data; Data preprocessing, using data cleaning technology to remove noise data from the initial data, fill in missing values, and generate more sample data through data augmentation technology, and integrate to obtain a dataset to be analyzed; Multi-modal feature extraction, extracting text features from the dataset to be analyzed through natural language processing technology, extracting image features through computer vision technology, extracting social relationship features through social network analysis technology, and integrating the text features, image features, and social relationship features to obtain basic data; Construct a graph neural network, import the basic data into the graph neural network to capture the complex relationships and behavior patterns among multiple members; Construct a behavior analysis model, input the basic data and the complex relationships and behavior patterns among multiple members into the behavior analysis model, and optimize the behavior analysis model through a multi-task learning framework, self-supervised learning technology, and Bayesian optimization method; Construct a time series prediction model, and combine the data obtained from the behavior analysis model with the time series prediction model to model the member behavior sequence and predict future behavior trends; Push personal recommendations, and generate a mixed recommendation for each member through a content recommendation mechanism and a collaborative filtering recommendation mechanism based on the future behavior trends.
[0007] Further, the step of constructing a graph neural network, importing the basic data into the graph neural network to capture the complex relationships and behavior patterns among multiple members includes: Taking members and activities as nodes in the graph neural network, and taking the relationships among multiple members as edges to construct a member relationship graph; Using the graph neural network to model the member relationship graph, and through node feature aggregation and information transfer mechanisms, learning high-dimensional node embedding representations to capture the complex relationships and behavior patterns among members.
[0008] Further, the step of taking members and activities as nodes in the graph neural network, taking the relationships among multiple members as edges to construct a member relationship graph further includes: Constructing a member knowledge graph, mapping the basic information and interest preferences of new members into the knowledge graph to generate member nodes and associating relevant background knowledge nodes; Constructing an activity knowledge graph, mapping the description information and features of new activities into the knowledge graph to generate activity nodes and associating relevant background knowledge nodes; Fusing the member knowledge graph and activities to obtain a complete recommendation knowledge graph.
[0009] Further, the step of pushing personal recommendations, generating a mixed recommendation for each member through a content recommendation mechanism and a collaborative filtering recommendation mechanism based on the future behavior trends includes: Calculating the similarity between member nodes and activity nodes based on the node relationships in the recommendation knowledge graph; Combining the results of the behavior prediction model with the similarity in the recommendation knowledge graph to generate personalized recommendation content; Generate a recommendation list based on node similarity and behavior prediction results, and sort the recommended content in the recommendation list according to similarity.
[0010] Further, the step of constructing a time series prediction model, combining the data obtained from the behavior analysis model with the time series prediction model, and modeling the member behavior sequence to predict future behavior trends further includes: Behavior sequence data collection: Collect the behavior data of members at different times and format the behavior data into a time series form. Arrange each behavior data in chronological order to form a time series. Data preprocessing: Align the behavior data in time, fill in the missing data points using interpolation filling and mean filling, and segment the behavior sequence into subsequences of a fixed length. Construct an LSTM model: Design a multi-layer LSTM network, input the preprocessed behavior sequence data into the LSTM network, and generate predictions for future behaviors. Construct a Transformer model: Stack multiple encoder layers and multiple decoder layers to obtain an encoder and a decoder. Then convert the behavior sequence data into an embedding representation and add positional encoding to the embedding representation to combine and obtain the Transformer model.
[0011] Further, the step of constructing a behavior analysis model, inputting the basic data and the complex relationships and behavior patterns among multiple members into the behavior analysis model, and optimizing the behavior analysis model through a multi-task learning framework, self-supervised learning techniques, and Bayesian optimization methods includes: Construct a multi-task learning framework: Define a main task and auxiliary tasks, design a shared layer and task-specific layers, extract the features of the main task and auxiliary tasks through the task-specific layers, and output the prediction results of the tasks. Design self-supervised learning tasks: Design a node reconstruction task to reconstruct node features, and then design a contrastive learning task to enhance the discrimination ability of node representations. Define a hyperparameter space and an optimization objective, run Bayesian optimization to find the optimal hyperparameter combination, and optimize the behavior analysis model.
[0012] Further, the step of constructing a multi-task learning framework, defining a main task and auxiliary tasks, designing a shared layer and task-specific layers, and extracting the features of the main task and auxiliary tasks through the task-specific layers and outputting the prediction results of the tasks further includes: Define predicting the future behavior of members as the main task, and defining classifying members' interest preferences and predicting members' activity participation as auxiliary tasks. Design a shared layer, construct a multi-layer shared neural network layer, and use the neural network layer to extract common features in the member data; Design a task-specific layer, design a specific neural network layer for each task, extract features related to the task, and output the prediction result of the task; Design a loss function, weight and sum the losses of all tasks as the total loss function, and optimize the behavior analysis model through the total loss function.
[0013] The present invention also proposes an intelligent recommendation device for member behavior analysis, and the member behavior analysis device includes: A multi-source information acquisition module, configured to acquire the basic information and preferences of new members, as well as the description information and features of new activities; A data processing module, configured to perform data preprocessing and feature extraction on the multi-source information; A model training module, configured to design a graph neural network training and a multi-task learning framework, and generate a behavior analysis model according to the processed data; A knowledge graph module, configured to construct a knowledge graph of members and activities, including member nodes, activity nodes, and their associated background knowledge nodes; A recommendation generation module, which performs recommendation analysis and calculation based on the knowledge graph and generates a recommendation list.
[0014] The present invention also proposes an intelligent recommendation device for member behavior analysis. The intelligent recommendation device for member behavior analysis includes: a memory, a processor, and an intelligent recommendation program for member behavior analysis stored on the memory and executable on the processor. When the intelligent recommendation program for member behavior analysis is executed by the processor, the steps of the intelligent recommendation method for member behavior analysis described in any one of the above technical solutions are implemented.
[0015] The present invention also proposes a storage medium, on which there is an intelligent recommendation program for member behavior analysis. When the intelligent recommendation program for member behavior analysis is executed by a processor, the steps of the intelligent recommendation method for member behavior analysis described in any one of the above technical solutions are implemented.
[0016] The present invention provides an intelligent recommendation method, device, equipment, and storage medium for member behavior analysis. This method collects and integrates member data from complex sources through multi-source data integration and real-time processing, and combines multi-modal feature extraction. Then, by constructing a graph neural network and a behavior analysis model, it effectively predicts the behavior of members and generates corresponding recommended content in a timely manner for pushing. The present invention can push personal content for members in real time and effectively, and effectively improve work efficiency and member satisfaction. Description of the Drawings
[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the accompanying drawings required for the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can be obtained according to the processes shown in these drawings.
[0018] Figure 1 It is a schematic flowchart of the first embodiment in the intelligent recommendation method for member behavior analysis; Figure 2 It is a schematic flowchart of the second embodiment in the intelligent recommendation method for member behavior analysis; Figure 3 It is a schematic flowchart of the third embodiment in the intelligent recommendation method for member behavior analysis; Figure 4 It is a schematic diagram of the module structure of an embodiment in the intelligent recommendation device for member behavior analysis; Figure 5 It is a schematic diagram of the device structure of the hardware operating environment of an embodiment in the intelligent recommendation device for member behavior analysis. Detailed implementation manners
[0019] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.
[0020] It should be noted that all directional indications (such as up, down, left, right, front, back...) in the embodiments of the present invention are only used to explain the relative positional relationship and movement conditions between components in a specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indications will also change accordingly.
[0021] In addition, the descriptions involving "first", "second", etc. in the present invention are only for descriptive purposes, and cannot be understood as indicating or implying their relative importance or implicitly indicating the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include at least one of such features. In addition, the technical solutions between various embodiments can be combined with each other, but it must be based on the fact that those of ordinary skill in the art can implement it. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection required by the present invention.
[0022] Multi-source heterogeneous data refers to data from different sources, with different types and structures. These data may have significant differences in format, content, semantics, storage methods, etc. Natural Language Processing (NLP) involved in this invention is a branch of artificial intelligence, aiming to enable computers to understand, interpret, and generate human language. NLP technologies include text analysis, sentiment analysis, entity recognition, and machine translation, etc. Computer Vision (CV) is also a field of artificial intelligence, aiming to enable computers to obtain useful information from images or videos. These computer vision technologies include image classification, object detection, face recognition, and image generation, etc. Social Network Analysis (SNA) is a method for studying social relationships and network structures, usually using graph theory and statistical methods to analyze and represent nodes (such as users) and edges (such as relationships between users) in social networks. Graph Neural Networks (GNNs) are a type of neural network specifically designed for processing graph-structured data. Through node feature aggregation and message passing mechanisms, they can learn high-dimensional node embedding representations and capture complex relationships between nodes.
[0023] Based on Figures 1-5 As shown, an intelligent recommendation method for member behavior analysis is disclosed in this invention, specifically including the following steps: S10. Collect member behavior data, integrate organizational activity records, and integrate multi-source heterogeneous data of members' data on the online learning platform and members' platform social interactions, and then perform data integration through the ETL process to obtain initial data; In an embodiment of this invention, collecting member behavior data can be to collect detailed records of members' participation in organizational activities, including data such as activity names, times, locations, participation status, etc. At the same time, it can also collect members' social interaction data, such as collecting data on members' interactions on social platforms, including friend relationships, messages, likes, comments, etc. After collecting multi-source heterogeneous data, then use the ETL (Extract, Transformer, Load, extraction, transformation, loading) process for data integration.
[0024] It is understandable that the extraction in the ETL process refers to extracting data from multi-source heterogeneous data sources, and then cleaning and transforming the extracted data to ensure the consistency of data formats and structures. The transformation process includes data cleaning (such as removing duplicate data, handling missing values, etc.), data merging, data summarization, and data format conversion, etc., to transform the original data into a data format and structure that meets the requirements of the target data warehouse. Loading means loading the transformed data into a data warehouse or a data storage system, and the loading process needs to ensure that the data is stored in the target system in an appropriate form and structure for easy querying and analysis.
[0025] Specifically, in this embodiment, the Apache Kafka stream processing technology is used to achieve real-time data collection and processing to ensure the timeliness of members' behavior data. Among them, the Apache Kafka stream processing technology is a distributed stream processing platform used to process a large number of real-time data streams, supporting high throughput, scalability, and persistence. More specifically, Kafka organizes members' behavior data into topics, and each topic is divided into multiple partitions, and the partitions enable Kafka to scale horizontally to handle higher throughput.
[0026] In another embodiment, the step of collecting members' behavior data can also collect members' learning data on various learning platforms, such as course completion status, learning duration, test scores, etc.
[0027] S20. Data preprocessing, using data cleaning technology to remove noise data from the initial data, fill in missing values, and generate more sample data through data augmentation technology, and integrating to obtain the dataset to be analyzed; In this embodiment, data cleaning technology is used to remove noise data from members' behavior data, fill in missing values, and standardize the data format, and more sample data can also be generated through data augmentation technology to improve the robustness of the model.
[0028] S30. Multi-modal feature extraction, using natural language processing technology to extract text features from the dataset to be analyzed, computer vision technology to extract image features, and social network analysis technology to extract social relationship features, and integrating the text features, image features, and social relationship features to obtain basic data; In this embodiment, by extracting features from text data (such as activity descriptions, learning notes, meeting records, etc.), useful information is extracted using techniques such as word embedding and named entity recognition (NER). Features are extracted from picture data (such as activity photos, promotional posters, etc.), and visual features are extracted through techniques such as convolutional neural networks (GNN). Social network graphs are used for social relationship data to analyze the interaction relationships between members.
[0029] S40. Construct a graph neural network, import the basic data into the graph neural network, and capture the complex relationships and behavior patterns among multiple members; S41. Use members and activities as nodes in the graph neural network, and use the relationships among multiple members as edges to construct a member relationship graph; S41a. Construct a member knowledge graph, map the basic information and interest preferences of new members into the knowledge graph, generate member nodes, and associate relevant background knowledge nodes; S41b. Construct an activity knowledge graph, map the description information and features of new activities into the knowledge graph, generate activity nodes, and associate relevant background knowledge nodes; S41c. Integrate the member knowledge graph and activities to obtain a complete recommendation knowledge graph; Specifically, in this embodiment, members and activities are used as nodes in the member relationship graph. The node features include the basic information of members, the characteristics of participating in organizational construction activities, etc. The relationships among members (such as friendship relationships, jointly participated activities, etc.) are used as edges in the member relationship graph, and the weights of the edges represent the intensity or frequency of the relationships. Use graph neural networks (GNNs) to model the member relationship graph, and through node feature aggregation and message passing mechanisms, learn high-dimensional node embeddings to represent the complex relationships and behavior patterns among members.
[0030] It can be understood that the Cold Start Problem is one of the main problems faced by recommendation systems, which refers to the lack of sufficient user behavior data in the initial stage of the system, making it difficult to accurately make recommendations. In the present invention, by adopting the Knowledge Graph technology, the basic information of members, activity information, and background knowledge are associated. The basic information and interest preferences of members are mapped into the knowledge graph to generate member nodes and associate the member nodes with relevant background knowledge nodes. Then, the description information and features of new organizational construction activities are mapped into the knowledge graph to generate activity nodes and associate relevant background knowledge nodes. Finally, the member knowledge graph and the activity knowledge graph are integrated to form a complete recommendation knowledge graph.
[0031] In this embodiment, the graph neural network model is implemented using a graph convolutional network (GCN, Graph Convolutional Network), which performs convolutional operations on each node and its neighbors to aggregate the features of neighbor nodes.
[0032] The specific formula is as follows:
[0033] Among them, is the The node features of the layer, A is the adjacency matrix, and D is the degree matrix. is the weight matrix of the layer, and is the activation function.
[0034] S42. Use a graph neural network to model the membership graph. Through the node feature aggregation and information transfer mechanism, learn the high-dimensional node embedding representation to capture the complex relationships and behavior patterns among members. It can be understood that the high-dimensional node embedding representation means representing the nodes in the membership graph as vectors in a high-dimensional space. Such vectors can capture the features and relationships of the nodes in the graph structure. In this way, the graph structure information is converted into a numerical form, enabling the use of various machine learning and deep learning algorithms for processing and analysis.
[0035] In this embodiment, the high-dimensional node embedding representation is generated through a graph convolutional network. In another embodiment, it is also possible to use random walks to generate node sequences, and then perform word embedding on these sequences to obtain the node embedding representation, as follows:
[0036] where D is the random walk to generate node pairs, and is the conditional probability.
[0037] S50. Build a behavior analysis model, input the basic data and the complex relationships and behavior patterns among multiple members into the behavior analysis model, and optimize the behavior analysis model through a multi-task learning framework, self-supervised learning technology, and Bayesian optimization method. S51. Build a multi-task learning framework, define the main task and auxiliary tasks, and design the shared layer and task-specific layers. Extract the features of the main task and auxiliary tasks through the task-specific layers and output the prediction results of the tasks. S51a. Define predicting the future behavior of members as the main task, and define classifying members' interest preferences and predicting members' activity participation as auxiliary tasks. S51b. Design the shared layer, build a multi-layer shared neural network layer, and use the neural network layer to extract the common features in the member data. S51c. Design the task-specific layers, design specific neural network layers for each task, extract the features related to the task, and output the prediction results of the task. S51d. Design the loss function, weight and sum the losses of all tasks as the total loss function, and optimize the behavior analysis model through the total loss function. In this embodiment, designing a multi-task learning framework can optimize multiple objective functions simultaneously, such as engagement prediction, interest preference prediction, etc., to improve the comprehensive performance of the model.
[0038] Specifically, the steps for designing the loss function are to design a suitable loss function, such as cross-entropy loss, contrastive loss, etc., optimize and adjust according to the specific task, and then use the gradient descent algorithm (such as the Adam optimizer) to train the model, update the weights through the backpropagation algorithm, and minimize the loss function.
[0039] S52. Design self-supervised learning tasks, design node reconstruction tasks to reconstruct node features, and then design contrastive learning tasks to enhance the discriminative ability of node representations; In this embodiment, the self-supervised learning tasks include contrastive learning and generative adversarial networks. Among them, contrastive learning is to enable the behavior analysis model to autonomously learn the latent structures and patterns in the data through contrastive learning techniques, enhancing the generalization ability. Specifically, it is to perform contrastive learning using positive and negative samples. The generative adversarial network is to use GANs to generate high-quality synthetic samples through the adversarial training of the generator and discriminator, and then further optimize the robustness and generalization ability of the behavior analysis model through these synthetic samples.
[0040] S53. Define the hyperparameter space and the optimization objective, run Bayesian optimization to find the optimal hyperparameter combination, and optimize the behavior analysis model; S60. Construct a time series prediction model, combine the data obtained from the behavior analysis model with the time series prediction model, and model the member behavior sequence to predict the future behavior trend; S61. Collect behavior sequence data, collect the behavior data of members at different times and format the behavior data into a time series form, arrange each behavior data in chronological order to form a time series; S62. Data preprocessing, perform time alignment on the behavior data, fill in the missing data points using interpolation filling and mean filling, and divide the behavior sequence into subsequences of a fixed length; S63. Construct an LSTM model, design a multi-layer LSTM network, input the preprocessed behavior sequence data into the LSTM network, and generate predictions for future behaviors; In this embodiment, LSTM (Long Short Term Memory networks) is a special type of RNN network (Recurrent Neural Networks). The RNN network can continuously cycle information to ensure the continuous existence of information, thus solving the problem of information loss that occurs when neural networks predict sequence information. The core of LSTM is the cell state. The repeating unit of LSTM is different from the unit in the standard RNN network which has only one network layer. It controls the cell state through the forget gate, input gate, and output gate. In this embodiment, the LSTM network architecture is obtained by constructing an input layer, an LSTM layer, and an output layer. The input layer is used to input the preprocessed behavioral sequence data into the LSTM network, and the input data format is (batch_size, sequence_length, feature_dimension). The LSTM layer specifically designs a multi-layer LSTM network. Each layer of the LSTM network unit consists of multiple memory cells, and a single memory cell controls the flow of information through the input gate, forget gate, and output gate to capture long-term and short-term dependencies. The output layer connects the output of the LSTM layer to a fully connected layer to generate predictions for future behaviors.
[0041] The specific formula is as follows:
[0042]
[0043]
[0044]
[0045]
[0046]
[0047] Among them, is the forget gate, is the input gate, is the candidate memory, is the current memory, is the output gate is the current hidden state, is the sigmoid activation function, tanh is the tanh activation function, and are the weights and configuration parameters.
[0048] S64. Construct a Transformer model by stacking multiple encoder layers and multiple decoder layers to obtain an encoder and a decoder. Then, convert the behavior sequence data into an embedding representation and add positional encoding to the embedding representation to combine and obtain the Transformer model; In this embodiment, the architecture of the Transformer model consists of an input embedding layer, an encoder, and a decoder. Among them, the input embedding layer converts the behavior sequence data into an embedding representation and adds positional encoding (Positional Encoding) to retain the time series information. The encoder (Encoder) is stacked by multiple encoder layers, and each encoder layer consists of a self-attention mechanism (Self-Attention) and a feed-forward neural network (Feed-Forward Neural Network). The decoder (Decoder) is stacked by multiple decoder layers, and each decoder layer includes a self-attention mechanism, an encoder-decoder attention mechanism, and a feed-forward neural network.
[0049] Specifically, the calculation formula of the self-attention mechanism is:
[0050]
[0051]
[0052]
[0053] Among them, Q, K, and V are the linear transformations of the query vector, key vector, and value vector respectively, 、 、 are weight matrices, is the dimension of the key vector.
[0054] The calculation formula of the positional encoding is:
[0055]
[0056] Among them, is the position, i is the dimension index, is the embedding dimension.
[0057] S70. Push personal recommendations, and generate a mixed recommendation for each member through a content recommendation mechanism and a collaborative filtering recommendation mechanism for future behavior trends; S71. Calculate the similarity between the member node and the activity node based on the node relationships in the recommendation knowledge graph; S72. Combine the results of the behavior prediction model with the similarity in the recommendation knowledge graph to generate personalized recommendation content; S73. Generate a recommendation list based on the node similarity and the behavior prediction results, and sort the recommendation content in the recommendation list according to the similarity.
[0058] As Figure 4 shown, the present invention also proposes an intelligent recommendation device for member behavior analysis. The member behavior analysis device includes: A multi-source information acquisition module 10 for acquiring the basic information and preferences of new members and the description information and features of new activities; A data processing module 20 for performing data preprocessing and feature extraction on the multi-source information; A model training module 30 for designing a graph neural network training and a multi-task learning framework, and generating a behavior analysis model according to the processed data; A knowledge graph module 40 for constructing a knowledge graph of members and activities, including member nodes, activity nodes, and their associated background knowledge nodes; A recommendation generation module 50 for performing recommendation analysis and calculation based on the knowledge graph to generate a recommendation list.
[0059] The intelligent recommendation device for member behavior analysis provided by the present invention adopts the intelligent recommendation method for member behavior analysis in the above embodiment, and can solve technical problems. Compared with the prior art, the beneficial effects of the intelligent recommendation device for member behavior analysis provided by the present invention are the same as those of the intelligent recommendation method for member behavior analysis provided by the above embodiment, and other technical features in the intelligent recommendation device for member behavior analysis are the same as the features disclosed in the method of the above embodiment, and will not be described again here.
[0060] The present invention also proposes an intelligent recommendation device for member behavior analysis. The intelligent recommendation device for member behavior analysis includes: a memory, a processor, and an intelligent recommendation program for member behavior analysis stored on the memory and executable on the processor. When the intelligent recommendation program for member behavior analysis is executed by the processor, it implements the steps of the intelligent recommendation method for member behavior analysis in any one of the above technical solutions.
[0061] Refer to Figure 5 , which shows a schematic structural diagram of an intelligent recommendation data integration device for member behavior analysis suitable for implementing one embodiment of the present invention.
[0062] The intelligent recommendation device for member behavior analysis in the embodiments of this application may include, but is not limited to, mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistant), PADs (Portable Application Description: tablet computers), PMPs (Portable Media Player), vehicle terminals (such as in-vehicle navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 3 The illustrated intelligent recommendation device for member behavior analysis is merely an example and should not impose any limitations on the functions and usage scope of the embodiments of this application.
[0063] As Figure 3 shown, the intelligent recommendation device for member behavior analysis may include a processing device 1001 (such as a central processing unit, a graphics processing unit, etc.), which may perform various appropriate actions and processes according to the program stored in the read-only memory (ROM: Read Only Memory) 1002 or the program loaded from the storage device 1003 into the random access memory (RAM: Random Access Memory) 1004. In the RAM 1004, various programs and data required for the operation of the intelligent recommendation device for member behavior analysis are also stored. The processing device 1001, the ROM 1002, and the RAM 1004 are connected to each other through a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Generally, the following systems may be connected to the I / O interface 1006: an input device 1007 including, for example, a touch screen, a touchpad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; an output device 1008 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage device 1003 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 1009. The communication device 1009 may allow the intelligent recommendation device for member behavior analysis to communicate with other devices wirelessly or wiredly to exchange data. Although the figure shows an intelligent recommendation device for member behavior analysis with various systems, it should be understood that it is not required to implement or have all the shown systems. Instead, more or fewer systems may be implemented or had.
[0064] In particular, according to the embodiments disclosed in the present application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, the embodiments disclosed in the present application include a computer program product that includes a computer program carried on a computer-readable medium, and the computer program contains program codes for executing the methods shown in the flowcharts. In such an embodiment, the computer program can be downloaded and installed from a network through a communication device, or installed from a storage device 1003, or installed from a ROM 1002. When the computer program is executed by a processing device 1001, the above functions defined in the methods of the embodiments disclosed in the present application are executed.
[0065] The intelligent recommendation device for member behavior analysis provided by the present application adopts the intelligent recommendation method for member behavior analysis in the above embodiments, and can solve the technical problems of intelligent recommendation for member behavior analysis. Compared with the prior art, the beneficial effects of the intelligent recommendation device for member behavior analysis provided by the present application are the same as those of the intelligent recommendation method for member behavior analysis provided by the above embodiments, and other technical features in the intelligent recommendation device for member behavior analysis are the same as the features disclosed in the method of the previous embodiment, and will not be elaborated here.
[0066] It should be understood that the various parts disclosed in the present invention can be implemented by hardware, software, firmware or a combination thereof. In the description of the above embodiments, specific features, structures, materials or characteristics can be combined in a suitable manner in any one or more embodiments or examples.
[0067] As described above, only the specific implementation manners of the present invention are provided, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed in the present application, and all should be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.
[0068] The present invention also proposes a storage medium with an intelligent recommendation program for member behavior analysis. When the intelligent recommendation program for member behavior analysis is executed by a processor, the steps of the intelligent recommendation method for member behavior analysis in any one of the above technical solutions are performed.
[0069] The storage medium provided by the present application is a computer-readable storage medium, having computer-readable program instructions (i.e., computer programs) stored thereon, and the computer-readable program instructions are used to execute the intelligent recommendation method for member behavior analysis in the above embodiments.
[0070] The computer-readable storage medium provided by the present invention can be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or components, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections with one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM) or flash memory, optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the above. In this embodiment, the computer-readable storage medium can be any tangible medium that contains or stores a program, which can be used by or in conjunction with an instruction execution system, device, or component. The program code contained on the computer-readable storage medium can be transmitted by any appropriate medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination of the above.
[0071] The above computer-readable storage medium can be included in the intelligent recommendation device for member behavior analysis; it can also exist independently without being assembled into the intelligent recommendation device for member behavior analysis.
[0072] Combining all of the above technical solutions, an intelligent recommendation method, device, equipment, and storage medium for member behavior analysis provided by the present invention collects and integrates member data from complex sources by introducing multi-source data integration and real-time processing, and combines multi-modal feature extraction. Then, by constructing a graph neural network and a behavior analysis model, it effectively predicts the behavior of members and generates corresponding recommended content in chronological order for pushing. The present invention can analyze the behavior data and behavior patterns of members in real time and effectively, judge the needs of members, and then generate personalized content pushes for members, which can effectively improve the work efficiency and member satisfaction of members, and provide a new technical means for organizational construction work.
[0073] The above are only the preferred embodiments of the present invention, and do not limit the patent scope of the present invention. Any equivalent structural transformation made under the inventive concept of the present invention using the content of the specification and drawings of the present invention, or any direct / indirect application in other related technical fields, is included in the patent protection scope of the present invention.
Claims
1. An intelligent recommendation method for member behavior analysis, characterized in that, It includes the following steps: Collect member behavior data, integrate organizational activity records, integrate multi-source heterogeneous data of members' data on the online learning platform and members' platform social interactions, and then perform data integration through the ETL process to obtain initial data; Data preprocessing, using data cleaning technology to remove noise data from the initial data, fill in missing values, and generate more sample data through data augmentation technology, and comprehensively obtain the dataset to be analyzed; Multi-modal feature extraction, extracting text features from the dataset to be analyzed through natural language processing technology, extracting image features through computer vision technology, and extracting social relationship features through social network analysis technology, and integrating the text features, image features, and social relationship features to obtain basic data; Construct a graph neural network, import the basic data into the graph neural network to capture the complex relationships and behavior patterns among multiple members; Construct a behavior analysis model, input the basic data and the complex relationships and behavior patterns among multiple members into the behavior analysis model, and optimize the behavior analysis model through a multi-task learning framework, self-supervised learning technology, and Bayesian optimization method; Construct a time series prediction model, and combine the data obtained from the behavior analysis model with the time series prediction model to model the member behavior sequence and predict future behavior trends; Push personal recommendations, and generate a mixed recommendation for each member through a content recommendation mechanism and a collaborative filtering recommendation mechanism based on the future behavior trends; 2. The intelligent recommendation method for member behavior analysis according to claim 1, wherein The step of constructing a graph neural network, importing the basic data into the graph neural network to capture the complex relationships and behavior patterns among multiple members, includes: Taking members and activities as nodes in the graph neural network, and taking the relationships among multiple members as edges to construct a member relationship graph; Using the graph neural network to model the member relationship graph, and learning high-dimensional node embedding representations through node feature aggregation and information transfer mechanisms to capture the complex relationships and behavior patterns among members; 3. The intelligent recommendation method for member behavior analysis according to claim 2, characterized in that, The step of taking members and activities as nodes in the graph neural network, and taking the relationships among multiple members as edges to construct a member relationship graph further includes: Constructing a member knowledge graph, mapping the basic information and interest preferences of new members into the knowledge graph to generate member nodes, and associating relevant background knowledge nodes; Constructing an activity knowledge graph, mapping the description information and features of new activities into the knowledge graph to generate activity nodes, and associating relevant background knowledge nodes; Fusing the member knowledge graph and activities to obtain a complete recommendation knowledge graph; 4. The intelligent recommendation method for member behavior analysis according to claim 3, wherein, The step of pushing personal recommendations, and generating a mixed recommendation for each member through a content recommendation mechanism and a collaborative filtering recommendation mechanism based on the future behavior trends, includes: Calculating the similarity between member nodes and activity nodes based on the node relationships in the recommendation knowledge graph; Combining the results of the behavior prediction model with the similarity in the recommendation knowledge graph to generate personalized recommendation content; Generating a recommendation list based on the node similarity and behavior prediction results, and sorting the recommendation content in the recommendation list according to the similarity.
5. The intelligent recommendation method for member behavior analysis according to claim 1, wherein, The steps of constructing a time series prediction model, combining the data obtained from the behavior analysis model with the time series prediction model, and modeling the member behavior sequence to predict future behavior trends further include: Collecting behavior sequence data, collecting the behavior data of members at different times, formatting the behavior data into a time series form, and arranging each behavior data in chronological order to form a time series; Data preprocessing, performing time alignment on the behavior data, filling in the missing data points using interpolation filling and mean filling, and splitting the behavior sequence into subsequences of a fixed length; Constructing an LSTM model, designing a multi-layer LSTM network, inputting the preprocessed behavior sequence data into the LSTM network, and generating predictions for future behaviors; Constructing a Transformer model, stacking multiple encoder layers and multiple decoder layers to obtain an encoder and a decoder, then converting the behavior sequence data into an embedded representation and adding positional encoding to the embedded representation to combine and obtain the Transformer model.
6. The intelligent recommendation method for member behavior analysis according to claim 1, wherein The steps of constructing the behavior analysis model, inputting the basic data, the complex relationships and behavior patterns among multiple members into the behavior analysis model, and optimizing the behavior analysis model through a multi-task learning framework, self-supervised learning techniques, and Bayesian optimization methods include: Constructing a multi-task learning framework, defining the main task and auxiliary tasks, designing a shared layer and task-specific layers, and extracting the features of the main task and auxiliary tasks through the task-specific layers and outputting the prediction results of the tasks; Designing self-supervised learning tasks, designing a node reconstruction task to reconstruct node features, and then designing a contrastive learning task to enhance the discrimination ability of node representations; Defining the hyperparameter space and the optimization objective, running Bayesian optimization to find the optimal combination of hyperparameters, and optimizing the behavior analysis model.
7. The intelligent recommendation method for member behavior analysis according to claim 6, wherein The steps of constructing the multi-task learning framework, defining the main task and auxiliary tasks, designing a shared layer and task-specific layers, and extracting the features of the main task and auxiliary tasks through the task-specific layers and outputting the prediction results of the tasks further include: Defining the prediction of the future behavior of members as the main task, and defining the classification of members' interest preferences and the prediction of members' activity participation as auxiliary tasks; Designing a shared layer, constructing a multi-layer shared neural network layer, and using the neural network layer to extract the common features in the member data; Designing task-specific layers, designing specific neural network layers for each task, extracting the features related to the task, and outputting the prediction results of the task; Designing a loss function, weighting and summing the losses of all tasks as the total loss function, and optimizing the behavior analysis model through the total loss function.
8. An intelligent recommendation device for member behavior analysis, characterized in that, The member behavior analysis device includes: A multi-source information acquisition module, used to acquire the basic information and preferences of new members, as well as the description information and features of new activities; A data processing module, used to perform data preprocessing and feature extraction on the multi-source information; A model training module, used to design graph neural network training and a multi-task learning framework, and generate a behavior analysis model based on the processed data; A knowledge graph module for constructing a knowledge graph of members and activities, including member nodes, activity nodes, and their associated background knowledge nodes; A recommendation generation module that performs recommendation analysis and calculation based on the knowledge graph to generate a recommendation list.
9. An intelligent recommendation device for member behavior analysis, characterized in that The intelligent recommendation device for member behavior analysis includes: a memory, a processor, and an intelligent recommendation program for member behavior analysis stored on the memory and executable on the processor. When the intelligent recommendation program for member behavior analysis is executed by the processor, it implements the steps of the intelligent recommendation method for member behavior analysis according to any one of claims 1-7.
10. A storage medium, characterized in that, There is an intelligent recommendation program for member behavior analysis on the storage medium. When the intelligent recommendation program for member behavior analysis is executed by the processor, it performs the steps of the intelligent recommendation method for member behavior analysis according to any one of claims 1-7.