Marketing method, system and equipment based on user portraits and medium

Abstract

The invention discloses a marketing method, system and device based on user portraits and a medium, and belongs to the technical field of marketing, and the method comprises the steps: obtaining behavior data of a plurality of users, carrying out the preprocessing operation, obtaining the preprocessed behavior data, carrying out the attribute feature extraction and probability prediction based on a user attribute model, constructing the user portraits of the plurality of users, and obtaining the user portraits of the plurality of users. Grouping to obtain a plurality of user groups; the method comprises the following steps: constructing a knowledge graph, extracting high-order neighbor node features based on a graph algorithm, generating a model by using marketing schemes to obtain a plurality of marketing schemes, inputting the knowledge graph and the plurality of marketing schemes into a recommendation model, and then selecting the marketing scheme corresponding to the highest recommendation degree to push to a corresponding user. According to the method, the multi-dimensional features of the user are fully processed through the user attribute model, the user portrait comprises multiple attribute features, the problem of single user portrait data is solved, the marketing scheme with the highest recommendation degree is selected and pushed to the client, and the marketing scheme recommendation accuracy is greatly improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of marketing, and particularly relates to a marketing method, system, device and medium based on user portraits. Background Art

[0002] With the rapid development of Internet technology and big data, the modern information volume has increased explosively, greatly enriching people's lives. However, it has also made it more time-consuming and difficult for people to obtain the required information from the data. With the growth of the user volume and interaction data, traditional data processing tools are difficult to efficiently process massive and high-dimensional user data. Moreover, various information of users is relatively fragmented, which is distributed in different platforms or systems, and the problem of data islands is likely to occur. With the growth of the user volume and interaction data, traditional data processing tools are difficult to efficiently process massive and high-dimensional user data and cannot quickly perform in-depth analysis.

[0003] Currently, the common methods for constructing user portraits based on user data include the method based on data statistics, the method based on machine learning, and the method based on deep learning. However, for relatively complex data distributions, the accuracy of feature extraction by the data statistics method will affect the results. When the methods based on machine learning and deep learning process user data, they can only mine features in a single dimension and ignore features in other dimensions, resulting in the inability to fully utilize the information in user data. Existing marketing systems generate marketing plans based on user personalized data and push marketing plans. Since user personalized data is relatively single, the user portraits generated based on user personalized data may have the problem of single data, resulting in a low accuracy of marketing plan recommendations. Therefore, there is an urgent need to provide a marketing method based on user portrait solutions to solve the problems of insufficient utilization of multi-dimensional features and low recommendation accuracy. Summary of the Invention

[0004] The purpose of the present invention is to provide a marketing method, system, device and medium based on user portraits to solve the above problems existing in the prior art.

[0005] To achieve the above purpose, the present invention adopts the following technical solutions: In a first aspect, the present invention provides a marketing method based on user portraits, including: Obtaining the behavior data of multiple users, where the behavior data includes behavior log data of cross-domain platforms, UGC content, and social relationship graphs, and performing a preprocessing operation on the behavior data to obtain preprocessed behavior data; Input the preprocessed behavior data into the user attribute model for attribute feature extraction and prediction to obtain the user's attribute features and the user's attribute probabilities. Based on the user's attribute features and the user's attribute probabilities, construct user portraits of multiple users, group the user portraits of multiple users, and obtain multiple user groups; Obtain the project portrait and the interaction relationship between the user and the project. Based on multiple user groups, the project portrait, and the interaction relationship between the user and the project, construct a knowledge graph. Among them, use the user portraits and the project portrait in the user group as the entity nodes of the knowledge graph, and use the interaction relationship between the user and the project as the edge of the knowledge graph; Find the adjacent nodes of each entity node in the knowledge graph, perform feature extraction on the adjacent nodes based on the graph algorithm to obtain high-order neighbor node features, input the user portrait and the high-order neighbor node features into the marketing plan generation model to obtain multiple marketing plans, and input the knowledge graph and multiple marketing plans into the recommendation model to obtain a recommendation degree sequence, and select the marketing plan corresponding to the first recommendation degree in the recommendation degree sequence and push it to the corresponding user.

[0006] In a possible design, perform preprocessing operations on the behavior data to obtain preprocessed behavior data, including: Perform data cleaning on the behavior data, and fill in the missing values of the behavior data after data cleaning to obtain complete behavior data; Perform moving average denoising and normalization on the complete behavior data to obtain preprocessed behavior data.

[0007] In a possible design, the user attribute model includes a pre-trained encoding layer, a DPCNN channel layer, a GRU and attention mechanism channel layer, and an output layer; the step of inputting the preprocessed behavior data into the user attribute model for attribute feature extraction and prediction to obtain the user's attribute features and the user's attribute probabilities includes: Input the preprocessed behavior data into the pre-trained encoding layer to obtain the data deep semantic information, and represent the data deep semantic information in a vectorized form to obtain the first data feature vector; Input the first data feature vector into the DPCNN channel layer for feature extraction to obtain the second data feature vector; Input the second data feature vector into the GRU and attention mechanism channel layer to obtain the first feature data; The output layer splices and fuses the first data feature vector, the second data feature vector, and the first feature data to obtain the user's attribute features, and uses the sigmoid function to predict the user's attribute features to obtain the user's attribute probabilities.

[0008] In a possible design, the step of inputting the second data feature vector into the GRU and attention mechanism channel layer to obtain the first feature data includes: Capture the long - distance dependencies in the second data feature vector based on the GRU model; Use the attention mechanism layer to calculate the weights of the captured long - distance dependencies, obtaining long - distance dependencies with different weights; Extract the features of the long - distance dependencies with different weights to obtain the first feature data.

[0009] In a possible design, the calculation expressions for the user's attribute features and user's attribute probabilities are: ; In the above formula, represents the fused feature data, represents the first data feature vector, represents the second data feature vector, represents the first feature data, represents the user's attribute probability, represents the sigmoid function, represents the activation function, represents the weight, represents the bias parameter.

[0010] In a possible design, group the user portraits of multiple users to obtain multiple user groups, including: Randomly select the user portrait of a user as the initial cluster center; Calculate the distance from each user's user portrait to the initial cluster center, and assign each user's user portrait to the cluster with a distance less than the preset distance; Calculate the mean of the user portraits within each cluster, use each mean as the cluster center of each cluster, recalculate the distance from each user's user portrait to the cluster center, and re - assign the user portraits to the cluster with a distance less than the preset distance until the preset termination condition is reached, obtaining multiple user groups.

[0011] In a possible design, the recommendation model includes an embedding layer, an attention - embedding propagation layer, and a prediction layer; the process of inputting the knowledge graph and multiple marketing plans into the recommendation model to obtain a recommendation degree sequence includes: The embedding layer embeds the knowledge graph and multiple marketing plans into a low - dimensional vector space to obtain initial embedding vectors; Input the initial embedding vectors into the attention - embedding propagation layer for calculation to obtain marketing plan weight coefficients; In the prediction layer, use the graph neural network to calculate the knowledge graph and multiple marketing plan weight coefficients to obtain multiple recommendation degrees, and sort the multiple recommendation degrees to obtain a recommendation degree sequence.

[0012] In a second aspect, the present invention provides a marketing system based on user portraits, including: An acquisition module, configured to acquire the behavior data of multiple users. The behavior data includes behavior log data of cross-domain platforms, UGC content, and social relationship graphs, and perform preprocessing operations on the behavior data to obtain preprocessed behavior data; A feature module, configured to input the preprocessed behavior data into a user attribute model for attribute feature extraction and prediction, obtain the attribute features and attribute probabilities of the users, construct user portraits of multiple users based on the attribute features and attribute probabilities of the users, and group the user portraits of multiple users to obtain multiple user groups; A construction module, configured to acquire project portraits and the interaction relationships between users and projects, and construct a knowledge graph based on multiple user groups, project portraits, and the interaction relationships between users and projects. Among them, the user portraits and project portraits in the user groups are used as entity nodes of the knowledge graph, and the interaction relationships between users and projects are used as edges of the knowledge graph; A recommendation module, configured to find adjacent nodes of each entity node in the knowledge graph, perform feature extraction of the adjacent nodes based on graph algorithms to obtain high-order neighbor node features, input the user portraits and high-order neighbor node features into a marketing plan generation model to obtain multiple marketing plans, input the knowledge graph and multiple marketing plans into a recommendation model to obtain a recommendation degree sequence, and select the marketing plan corresponding to the first recommendation degree in the recommendation degree sequence and push it to the corresponding user.

[0013] In a third aspect, the present invention provides a computer device, including a memory, a processor, and a transceiver that are communicatively connected in sequence. Among them, the memory is used to store computer programs, the transceiver is used to send and receive messages, and the processor is used to read the computer programs and execute the marketing method based on user portraits as described in the first aspect.

[0014] In a fourth aspect, the present invention provides a computer-readable storage medium, on which instructions are stored. When the instructions are run on a computer, the marketing method based on user portraits as described in the first aspect is executed.

[0015] The beneficial effects of the present invention are as follows: The present invention discloses a marketing method, system, device and medium based on user portraits. First, the behavior data of multiple users is acquired and preprocessed to obtain preprocessed behavior data. An attribute feature extraction and prediction are performed using a user attribute model. User portraits of multiple users are constructed based on the attribute features and attribute probabilities of the users, and the user portraits are grouped to obtain multiple user groups. A project portrait and the interaction relationship between the user and the project are acquired. A knowledge graph is constructed based on the multiple user groups, the project portrait and the interaction relationship between the user and the project. High-order neighbor node features are obtained according to the knowledge graph. The user portraits and the high-order neighbor node features are input into a marketing plan generation model to obtain multiple marketing plans. A recommendation model is used to generate recommendation degrees and sort them to obtain a recommendation degree sequence. The marketing plan corresponding to the first recommendation degree is selected and pushed to the corresponding user. By fully processing the multi-dimensional features of the user through the user attribute model, user portraits are constructed, where the user portraits include multiple user attribute features, solving the problem of single user portrait data. A knowledge graph is constructed based on the user portraits and the project portraits, intuitively reflecting the preference relationship between the user and the project. After generating the marketing plans, the marketing plans are sorted, and the marketing plan with the highest recommendation degree is selected and pushed to the customer, greatly improving the recommendation accuracy of the marketing plans, achieving the maximization of marketing effects, reducing marketing costs, increasing the return on investment, and facilitating practical application and promotion. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a flowchart of the marketing method based on user portraits provided in the first aspect of this embodiment; Figure 2 It is a block diagram of the modules of the marketing system based on user portraits provided in the second aspect of this embodiment. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0017] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the present invention in combination with the drawings and the descriptions of the embodiments or the prior art. Obviously, the following descriptions of the structures of the drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts. It should be noted here that the descriptions of these embodiment modes are used to help understand the present invention, but do not constitute a limitation to the present invention.

[0018] It should be understood that although terms such as first and second may be used herein to describe various units, these units should not be limited by these terms. These terms are only used to distinguish one unit from another. For example, the first unit can be called the second unit, and similarly, the second unit can be called the first unit, without departing from the scope of the exemplary embodiments of the present invention.

[0019] Embodiment: AsFigure 1 As shown in Figure 1 , in the first aspect of this embodiment, a marketing method based on user portraits is provided, including but not limited to the following steps: S1. Obtain the behavior data of multiple users. The behavior data includes behavior log data of cross-domain platforms, UGC content (User-Generated Content), and social relationship graphs. Perform preprocessing operations on the behavior data to obtain preprocessed behavior data; Specifically, in step S1, performing preprocessing operations on the behavior data to obtain preprocessed behavior data includes: S11. Perform data cleaning on the behavior data, and fill in the missing values of the behavior data after data cleaning to obtain complete behavior data; S12. Perform denoising and normalization on the complete behavior data using the moving average method to obtain preprocessed behavior data.

[0020] In this embodiment, data cleaning is performed on the behavior data, and the missing parts are filled to avoid the influence of outliers and missing values on the subsequent data processing process. And the moving average method is used to perform denoising and normalization operations on the complete behavior data to reduce the influence of noise and improve the robustness of the model; the principle of denoising by the moving average method is to remove the noise in the data by calculating the local average values of each data point in the data set.

[0021] S2. Input the preprocessed behavior data into the user attribute model for attribute feature extraction and prediction to obtain the attribute features and attribute probabilities of the users. Based on the attribute features and attribute probabilities of the users, construct user portraits of multiple users, and group the user portraits of multiple users to obtain multiple user groups; Furthermore, the user attribute model includes a pre-trained encoding layer, a DPCNN (Deep Pyramid Convolutional Neural Networks for Text Categorization) channel layer, a GRU (Gated Recurrent Unit) and attention mechanism channel layer, and an output layer.

[0022] Specifically, in step S2, inputting the preprocessed behavior data into the user attribute model for attribute feature extraction and prediction to obtain the attribute features and attribute probabilities of the users includes: S21. Input the preprocessed behavior data into the pre-trained encoding layer to obtain the deep semantic information of the data, and represent the deep semantic information of the data in a vectorized form to obtain the first data feature vector; S22. Input the first data feature vector into the DPCNN channel layer for feature extraction to obtain the second data feature vector; S23. Input the second data feature vector into the GRU and the attention mechanism channel layer to obtain the first feature data; Specifically, in step S23, inputting the second data feature vector into the GRU and the attention mechanism channel layer to obtain the first feature data includes: S231. Capture the long-distance dependence relationship in the second data feature vector based on the GRU model; Specifically, in step S231, two gating units are mainly used in the GRU model. The two gating units are mainly the update gate and the reset gate. Among them, the update gate is used to control how much historical information needs to be retained in the current state and how much new information is received, and the reset gate is used to determine how much information of the previous moment's network state is written. The specific calculation formulas of the update gate and the reset gate are as follows: ; In the above formula, represents the current input, represents the previous moment's state, represents the current moment's state, represents the activation function, represents the current learnable connection weight matrix of the update gate, represents the current learnable connection weight matrix of the reset gate, represents the previous moment's learnable connection weight matrix of the update gate, represents the previous moment's learnable connection weight matrix of the reset gate, represents the learnable offset vector of the update gate, represents the learnable offset vector of the reset gate.

[0023] S232. Use the attention mechanism layer to calculate the weights of the captured long-distance dependence relationships to obtain long-distance dependence relationships with different weights; Generally, the attention mechanism can be divided into the hard attention mechanism, the soft attention mechanism, and the self-attention mechanism. Preferably, in this embodiment, the attention mechanism is the self-attention mechanism, which can efficiently calculate the attention weights of all positions in parallel and learn the relative importance between each position and other positions, can effectively capture the relationships between different positions in the text sequence, making the neural network pay more attention to important features and suppressing the influence of irrelevant features.

[0024] S233. Extract the features of the long-distance dependence relationships with different weights to obtain the first feature data.

[0025] S24. The output layer splices and fuses the first data feature vector, the second data feature vector, and the first feature data to obtain the user's attribute features, and uses the sigmoid function to predict the user's attribute features to obtain the user's attribute probability.

[0026] Furthermore, the calculation expression of the user's attribute characteristics and the user's attribute probability is as follows: ; In the above formula, represents the fused feature data, represents the first data feature vector, represents the second data feature vector, represents the first feature data, represents the user's attribute probability, represents the sigmoid function, represents the activation function, represents the weight, represents the bias parameter.

[0027] Specifically, in step S2, the user portraits of multiple users are grouped to obtain multiple user groups, including: S25. Randomly select the user portrait of a user as the initial cluster center; S26. Calculate the distance from each user's user portrait to the initial cluster center, and assign each user's user portrait to the cluster with a distance less than the preset distance; S27. Calculate the mean value of the user portraits in each cluster, use each mean value as the cluster center of each cluster, calculate the distance from each user's user portrait to the cluster center again, and reassign the user portraits to the cluster with a distance less than the preset distance until the preset termination condition is reached to obtain multiple user groups.

[0028] S3. Obtain the project portrait and the interaction relationship between the user and the project, and construct a knowledge graph based on multiple user groups, the project portrait and the interaction relationship between the user and the project. Among them, the user portraits and project portraits in the user group are used as the entity nodes of the knowledge graph, and the interaction relationship between the user and the project is used as the edge of the knowledge graph; S4. Search for the adjacent nodes of each entity node in the knowledge graph, extract the features of the adjacent nodes based on the graph algorithm to obtain the high-order neighbor node features, input the user portrait and the high-order neighbor node features into the marketing plan generation model to obtain multiple marketing plans, input the knowledge graph and multiple marketing plans into the recommendation model to obtain a recommendation degree sequence, and select the marketing plan corresponding to the first recommendation degree in the recommendation degree sequence and push it to the corresponding user.

[0029] In step S4, the marketing plan generation model is a prior art. The recurrent neural network (RNN) and the Transformer can be selected to generate marketing plans, or the generative adversarial network can be used to generate marketing plans.

[0030] Further, the recommendation model includes an embedding layer, an attention embedding propagation layer, and a prediction layer.

[0031] Specifically, in step S4, the knowledge graph and multiple marketing plans are input into the recommendation model to obtain a recommendation degree sequence, including: S41. The embedding layer embeds the knowledge graph and multiple marketing plans into a low-dimensional vector space to obtain initial embedding vectors; Among them, the representation form of the knowledge graph is a triple (h, r, t), where h and t are entity nodes in the knowledge graph, r is the connection relationship in the knowledge graph (i.e., the edge in the knowledge graph), and multiple triples can share the same entity. The entity acts as the head or tail in different triples and serves as the association medium for different triples.

[0032] S42. The initial embedding vectors are input into the attention embedding propagation layer for calculation to obtain marketing plan weight coefficients; S43. In the prediction layer, a graph neural network is used to calculate the knowledge graph and multiple marketing plan weight coefficients to obtain multiple recommendation degrees, and the multiple recommendation degrees are sorted to obtain a recommendation degree sequence.

[0033] In this embodiment, a marketing method based on user portraits is provided, which includes obtaining the behavior data of multiple users. The behavior data includes behavior log data of cross-domain platforms, UGC content, and social relationship graphs. Perform preprocessing operations on the behavior data to obtain preprocessed behavior data; input the preprocessed behavior data into a user attribute model for attribute feature extraction and prediction to obtain the attribute features and attribute probabilities of the users. Based on the attribute features and attribute probabilities of the users, construct user portraits of multiple users, and group the user portraits of multiple users to obtain multiple user groups; obtain project portraits and the interaction relationships between users and projects, and construct a knowledge graph based on multiple user groups, project portraits, and the interaction relationships between users and projects. Among them, the user portraits and project portraits in the user groups are used as entity nodes of the knowledge graph, and the interaction relationships between users and projects are used as edges of the knowledge graph; find the adjacent nodes of each entity node in the knowledge graph, extract the features of the adjacent nodes based on graph algorithms to obtain high-order neighbor node features, input the user portraits and high-order neighbor node features into a marketing plan generation model to obtain multiple marketing plans, input the knowledge graph and multiple marketing plans into a recommendation model to obtain a recommendation degree sequence, and select the marketing plan corresponding to the first recommendation degree in the recommendation degree sequence and push it to the corresponding user. By extracting the attribute features and attribute probabilities of users through the user attribute model, constructing user portraits, and using a clustering algorithm to group the user portraits of multiple users, and constructing a knowledge graph according to the relationship between user groups and projects and between them, various types of data are effectively integrated, the characteristics of users and projects are comprehensively captured, the richness of the knowledge graph is enhanced, and the recommendation accuracy is improved; through the cooperation of the marketing plan generation model and the recommendation model, the marketing effect is maximized, the user experience is improved at the same time, and the user stickiness is enhanced.

[0034] As Figure 2 shown, in the second aspect of this embodiment, a marketing system based on user portraits is provided, including: An acquisition module for obtaining the behavior data of multiple users. The behavior data includes behavior log data of cross-domain platforms, UGC content, and social relationship graphs, and performing preprocessing operations on the behavior data to obtain preprocessed behavior data; A feature module for inputting the preprocessed behavior data into a user attribute model for attribute feature extraction and prediction to obtain the attribute features and attribute probabilities of the users, constructing user portraits of multiple users based on the attribute features and attribute probabilities of the users, and grouping the user portraits of multiple users to obtain multiple user groups; A construction module for obtaining project portraits and the interaction relationships between users and projects, and constructing a knowledge graph based on multiple user groups, project portraits, and the interaction relationships between users and projects. Among them, the user portraits and project portraits in the user groups are used as entity nodes of the knowledge graph, and the interaction relationships between users and projects are used as edges of the knowledge graph; A recommendation module is used to find adjacent nodes of each entity node in the knowledge graph, extract features of the adjacent nodes based on graph algorithms to obtain high-order neighbor node features, input the user profile and high-order neighbor node features into a marketing plan generation model to obtain multiple marketing plans, input the knowledge graph and the multiple marketing plans into a recommendation model to obtain a recommendation degree sequence, and select the marketing plan corresponding to the first recommendation degree in the recommendation degree sequence and push it to the corresponding user.

[0035] For the working process, working details and technical effects of the foregoing system provided in the second aspect of this embodiment, reference may be made to the marketing method described in the first aspect, which will not be elaborated herein.

[0036] In the third aspect of this embodiment, a computer device for executing the marketing method described in the first aspect is provided, which includes a memory, a processor, and a transceiver that are sequentially communicatively connected. Among them, the memory is used to store a computer program, the transceiver is used to send and receive messages, and the processor is used to read the computer program and execute the marketing method described in the first aspect. Specifically, for example, the memory may include, but is not limited to, a random access memory (RAM), a read-only memory (ROM), a flash memory, a first in first out memory (FIFO), and / or a first in last out memory (FILO), etc.; the processor may include, but is not limited to, a microprocessor of the STM32F105 series. In addition, the computer device may further include, but is not limited to, a power module, a display screen, and other necessary components.

[0037] For the working process, working details and technical effects of the foregoing computer device provided in the third aspect of this embodiment, reference may be made to the marketing method described in the first aspect, which will not be elaborated herein.

[0038] In the fourth aspect of this embodiment, a computer-readable storage medium storing instructions including the marketing method described in the first aspect is provided, that is, instructions are stored on the computer-readable storage medium, and when the instructions are run on a computer, the marketing method described in the first aspect is executed. Among them, the computer-readable storage medium refers to a carrier for storing data, and may include, but is not limited to, a floppy disk, an optical disk, a hard disk, a flash memory, a USB flash drive, and / or a memory stick, etc. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices.

[0039] For the working process, working details and technical effects of the foregoing computer-readable storage medium provided in the fourth aspect of this embodiment, reference may be made to the marketing method described in the first aspect, and details are not described herein again.

[0040] Finally, it should be noted that the above are only preferred embodiments of the present invention and are not intended to limit the protection scope of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A marketing method based on user portraits, characterized in that Including: Obtain the behavior data of multiple users, where the behavior data includes behavior log data, UGC content, and social relationship graphs across different platforms, and perform preprocessing operations on the behavior data to obtain preprocessed behavior data; Input the preprocessed behavior data into the user attribute model for attribute feature extraction and prediction to obtain the attribute features and attribute probabilities of the users. Based on the attribute features and attribute probabilities of the users, construct user portraits of multiple users, and group the user portraits of multiple users to obtain multiple user groups; Obtain the project portrait and the interaction relationship between the user and the project, and construct a knowledge graph based on multiple user groups, the project portrait, and the interaction relationship between the user and the project. Among them, the user portraits and project portraits in the user group are used as entity nodes of the knowledge graph, and the interaction relationship between the user and the project is used as the edge of the knowledge graph; Search for adjacent nodes of each entity node in the knowledge graph, perform feature extraction on the adjacent nodes based on graph algorithms to obtain high-order neighbor node features, input the user portrait and high-order neighbor node features into the marketing plan generation model to obtain multiple marketing plans, input the knowledge graph and multiple marketing plans into the recommendation model to obtain a recommendation degree sequence, and select the marketing plan corresponding to the first recommendation degree in the recommendation degree sequence and push it to the corresponding user.

2. The marketing method based on user portraits according to claim 1, wherein Perform preprocessing operations on the behavior data to obtain preprocessed behavior data, including: Perform data cleaning on the behavior data, and fill in missing values for the behavior data after data cleaning to obtain complete behavior data; Perform denoising and normalization on the complete behavior data using the moving average method to obtain preprocessed behavior data.

3. A marketing method based on user portraits according to claim 1, characterized in that The user attribute model includes a pre-trained encoding layer, a DPCNN channel layer, a GRU and attention mechanism channel layer, and an output layer; The step of inputting the preprocessed behavior data into the user attribute model for attribute feature extraction and prediction to obtain the attribute features and attribute probabilities of the users includes: Input the preprocessed behavior data into the pre-trained encoding layer to obtain deep semantic information of the data, and represent the deep semantic information of the data in a vectorized form to obtain the first data feature vector; Input the first data feature vector into the DPCNN channel layer for feature extraction to obtain the second data feature vector; Input the second data feature vector into the GRU and attention mechanism channel layer to obtain the first feature data; The output layer splices and fuses the first data feature vector, the second data feature vector, and the first feature data to obtain the attribute features of the user, and uses the sigmoid function to predict the attribute features of the user to obtain the attribute probability of the user.

4. A marketing method based on user portraits according to claim 3, characterized in that, Input the second data feature vector into the GRU and attention mechanism channel layer to obtain the first feature data, including: Capture the long-distance dependence relationship in the second data feature vector based on the GRU model; Use the attention mechanism layer to calculate the weights of the captured long-distance dependence relationship to obtain long-distance dependence relationships with different weights; Extract the features of the long-distance dependence relationships with different weights to obtain the first feature data.

5. The marketing method based on user portraits according to claim 3, characterized in that, The calculation expressions for the attribute features and attribute probabilities of the user are: ； In the above formula, represents the fused feature data, represents the first data feature vector, represents the second data feature vector, represents the first feature data, represents the attribute probability of the user, represents the sigmoid function, represents the activation function, represents the weight, represents the bias parameter.

6. The marketing method based on user portraits according to claim 1, characterized in that, Group the user portraits of multiple users to obtain multiple user groups, including: Randomly select the user profile of a user as the initial cluster center; Calculate the distance from the user profile of each user to the initial cluster center, and assign the user profile of each user to the cluster with a distance less than the preset distance; Calculate the mean value of the user profiles within each cluster, use each mean value as the cluster center of each cluster, calculate the distance from the user profile of each user to the cluster center again, and reassign the user profiles to the clusters with a distance less than the preset distance until the preset termination condition is reached, obtaining multiple user groups.

7. A marketing method based on user portraits according to claim 1, characterized in that The recommendation model includes an embedding layer, an attention embedding propagation layer, and a prediction layer; inputting the knowledge graph and multiple marketing plans into the recommendation model to obtain a recommendation degree sequence, including: The embedding layer embeds the knowledge graph and multiple marketing plans into a low-dimensional vector space to obtain initial embedding vectors; Input the initial embedding vectors into the attention embedding propagation layer for calculation to obtain marketing plan weight coefficients; In the prediction layer, use a graph neural network to calculate the knowledge graph and multiple marketing plan weight coefficients to obtain multiple recommendation degrees, and sort the multiple recommendation degrees to obtain a recommendation degree sequence.

8. A marketing system based on user portraits, characterized in that, Including: An acquisition module, configured to acquire the behavior data of multiple users, where the behavior data includes behavior log data of a cross-domain platform, UGC content, and a social relationship graph, and perform preprocessing operations on the behavior data to obtain preprocessed behavior data; A feature module, configured to input the preprocessed behavior data into a user attribute model for attribute feature extraction and prediction to obtain the attribute features and attribute probabilities of the users, construct the user profiles of multiple users based on the attribute features and attribute probabilities of the users, and group the user profiles of multiple users to obtain multiple user groups; A construction module, configured to acquire the project profile and the interaction relationship between the user and the project, and construct a knowledge graph based on the multiple user groups, the project profile, and the interaction relationship between the user and the project, where the user profiles and project profiles in the user group are used as entity nodes of the knowledge graph, and the interaction relationship between the user and the project is used as the edge of the knowledge graph; A recommendation module, configured to find the adjacent nodes of each entity node in the knowledge graph, perform feature extraction on the adjacent nodes based on a graph algorithm to obtain high-order neighbor node features, input the user profile and the high-order neighbor node features into a marketing plan generation model to obtain multiple marketing plans, input the knowledge graph and the multiple marketing plans into the recommendation model to obtain a recommendation degree sequence, and select the marketing plan corresponding to the first recommendation degree in the recommendation degree sequence and push it to the corresponding user.

9. A computer device, characterized in that, Including a memory, a processor, and a transceiver that are communicatively connected in sequence, where the memory is used to store a computer program, the transceiver is used to send and receive messages, and the processor is used to read the computer program and execute the user portrait-based marketing method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, Instructions are stored on the computer-readable storage medium, and when the instructions are run on a computer, the user portrait-based marketing method according to any one of claims 1 to 7 is executed.

Images