User portrait description method and system based on big data

Abstract

The invention discloses a user portrait description method and system based on big data, and belongs to the technical field of user portrait description, the method comprises the following steps: firstly, using a user entity, a commodity entity and a label entity as nodes to construct a heterogeneous information graph; secondly, obtaining user initial attribute information and label emotion attribute information; then, establishing an attribute propagation matrix according to the node connection relationship among the nodes; based on the attribute propagation matrix, multi-round attribute propagation is carried out, and iteration is carried out on propagation weights among nodes in each round of propagation so as to obtain user attribute information; and finally, user portrait description is completed by using the user attribute information, so that the user portrait can be applied to personalized recommendation. According to the method, the emotion features and the preference features of the user portrait are obtained by analyzing the entities and the node connection relations among the entities, so that deep description of the user portrait is completed, accurate user attribute information is obtained through multi-round attribute propagation, and the accuracy of the user portrait is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of user portrait characterization, and particularly relates to a method and system for user portrait characterization based on big data. Background Art

[0002] In the current digital wave, the user portrait, as a target user model constructed based on data, has become the core tool for enterprises to understand users and make precise policies. It comprehensively collects and deeply analyzes multi-dimensional information such as the demographic characteristics, behavior habits, hobby preferences, and consumption capabilities of users, and carefully creates a complete and three-dimensional set of user personality characteristics.

[0003] The emergence of the user portrait enables enterprises to understand user characteristics from multiple perspectives and in a refined manner. It can cover everything from basic demographic information such as age, gender, and education level, to the strength of consumption ability, the categories of consumption preferences, and daily hobbies and interests. This all-round characterization provides a solid basis for enterprises to formulate targeted marketing and product strategies. For example, for the young female group, enterprises can launch fashionable beauty products that meet their aesthetics and needs, and formulate personalized marketing plans, thereby improving user stickiness and optimizing the service experience. At the same time, the user portrait can also help enterprises keenly discover users' concerns, achieve precise marketing, and accurately push the right products or services to users with needs, greatly enhancing the marketing effect.

[0004] However, the existing user portrait characterization methods still have some deficiencies. First, most of the user portrait characterization methods are mainly based on user basic information and basic online operation data, and there is a serious lack of consideration for deep features such as user emotional preferences. This makes the finally constructed user portrait have a low correlation with the product and is difficult to accurately reflect the internal connection between the user and the product. Second, due to the lack of analysis of deep features, the accuracy of the user portrait is also greatly reduced, and it cannot provide truly effective decision-making support for enterprises.

[0005] Therefore, based on the above deficiencies, how to provide a method and system for user portrait characterization based on big data that can complete the deep characterization of the user portrait and improve the accuracy of the user portrait has become an urgent problem to be solved. Summary of the Invention

[0006] The purpose of the present invention is to provide a method and system for user portrait characterization based on big data to solve the problems existing in the prior art, namely, the low correlation between the user portrait and the product, and the insufficient accuracy of the user portrait.

[0007] To achieve the above purpose, the present invention adopts the following technical solutions: In a first aspect, the present invention provides a method for depicting user portraits based on big data, which includes: Based on a big data platform, obtain user entities, commodity entities, and label entities, and use the user entities, commodity entities, and label entities as nodes to construct a heterogeneous information graph, where the commodity entity is used to represent the interaction object with the user entity, and the label entity is used to represent the descriptive keywords defined by the user entity and the big data platform for the commodity entity; Based on the heterogeneous information graph, utilize the node connection relationship between the user entity and the commodity entity to obtain the initial user attribute information of the user entity, and utilize the node connection relationship between the commodity entity and the label entity to obtain the label sentiment attribute information of the label entity; According to the node connection relationship between each node in the heterogeneous information graph, establish an attribute propagation matrix; Input the initial user attribute information into a graph convolutional network, and based on the attribute propagation matrix, perform multiple rounds of attribute propagation and iterate the propagation weights between each node in each round to obtain user attribute information; Utilize the user attribute information to complete the depiction of the user portrait and apply the user portrait to personalized recommendation.

[0008] In a possible design, based on a big data platform, obtaining user entities, commodity entities, and label entities includes: Obtain the user registration information filled in by the user during registration, and based on the big data platform, obtain the user interaction behavior information of the user on the big data platform, and use the user registration information and the user interaction behavior information to form a user entity; Based on the big data platform, obtain the metadata information of the commodity, and use the metadata information of the commodity as the commodity entity; Based on the big data platform, obtain the evaluation information annotated by the user for the commodity, and combine the definition description of the commodity by the big data platform to form a label entity; Correspondingly, using the user entity, the commodity entity, and the label entity as nodes to construct a heterogeneous information graph includes: Define the user entity as a user node, define the commodity entity as a commodity node, define the label entity as a label node, and based on the big data platform, obtain the node connection relationship between the three types of nodes, where an edge connection is formed between the user node and the commodity node through a user evaluation relationship, an edge connection is formed between the commodity node and the label node through a definition description relationship, and an edge connection is formed between the user node and the label node through a keyword usage relationship; Based on the three types of nodes and the node connection relationship between the three types of nodes, construct a heterogeneous information graph.

[0009] In a possible design, based on the heterogeneous information graph, using the node connection relationship between the user entity and the commodity entity, the initial user attribute information of the user entity is obtained, including: Extract the basic user features in the user entity, where the basic user features include user age, user gender, user occupation, and user residence; Based on the node connection relationship between the user node and the commodity node, the commodity preference vectors of the user for different commodities are calculated respectively; Combine the basic user features and the commodity preference vectors of the user for different commodities to obtain the initial user attribute information of the user entity; Correspondingly, using the node connection relationship between the commodity entity and the label entity, the label sentiment attribute information of the label entity is obtained, including: Extract the evaluation behavior features in the label entity, where the evaluation behavior features include commodity score, evaluation type, and the proportion of each type of evaluation; Based on the node connection relationship between the commodity entity and the label entity, perform sentiment annotation on the keywords in the evaluation information labeled by the user for the commodity, so as to extract the sentiment preference vectors of the user for different labels; Combine the evaluation behavior features and the sentiment preference vectors of the user for different labels to obtain the label sentiment attribute information of the label entity.

[0010] In a possible design, performing sentiment annotation on the keywords in the evaluation information labeled by the user for the commodity includes: Obtain the label keywords in the label entity to perform semantic disambiguation on the label keywords to form unique label keywords; Obtain the NRC sentiment dictionary, and map the unique label keywords to the NRC sentiment dictionary through the semantic similarity algorithm; If there is a matching sentiment category between the unique label keyword and the NRC sentiment dictionary, count the occurrence frequency of the sentiment category, and select the sentiment category with the highest frequency as the sentiment-related word of the label entity; If there is no matching sentiment category between the unique label keyword and the NRC sentiment dictionary, for the unique label keyword, use the semantic similarity threshold screening method combined with context information to generate a sentiment-related word; Use the sentiment-related word to perform sentiment annotation on the keywords in the evaluation information labeled by the user for the commodity.

[0011] In a possible design, according to the node connection relationship between each node in the heterogeneous information graph, an attribute propagation matrix is established, including: According to the node connection relationships among the nodes in the heterogeneous information graph, initial propagation weights are assigned to each node connection relationship according to the node connection relationships among the three types of nodes. Among them, the initial propagation weights include the first-edge propagation weight, the second-edge propagation weight, and the third-edge propagation weight. Moreover, the first-edge propagation weight is the user's evaluation of the commodity, the second-edge propagation weight is the description frequency of the label for the commodity, and the third-edge propagation weight is the frequency of the user using the label; Based on the number of nodes in the heterogeneous information graph, a weight matrix is established. The initial propagation weights are filled into the weight matrix and normalized to use the weight matrix as the attribute propagation matrix.

[0012] In a possible design, the user's initial attribute information is input into a graph convolutional network. Based on the attribute propagation matrix, multiple rounds of attribute propagation are performed, including: Obtain the metadata information of the commodity, the user's initial attribute information, and the label sentiment attribute information as the initial attribute propagation information, and use the initial attribute propagation information as the input quantity to input into the graph convolutional network; For each node corresponding to a user entity, use the attribute propagation matrix to perform weighted aggregation on the attribute information of adjacent nodes to perform multiple rounds of attribute propagation; Correspondingly, iterating the propagation weights between each node in each round of propagation to obtain the user attribute information includes: In multiple rounds of attribute propagation, fuse the attribute propagation information in each round of attribute propagation with the initial attribute propagation information through a smoothing factor to complete the update of the attribute propagation information; Obtain a preset maximum number of iterations and a preset change rate threshold. When the change rate of the attribute propagation information is lower than the preset change rate threshold or reaches the maximum number of iterations, stop the attribute propagation to obtain the final attribute propagation information, and use the final attribute propagation information as the user attribute information. Among them, the user attribute information includes commodity preference, sentiment preference, and evaluation behavior.

[0013] In a possible design, using the user attribute information to complete the user portrait characterization and applying the user portrait to personalized recommendation includes: Obtain a personalized prediction model, and input the user portrait into the commodity preference prediction model to output a commodity preference prediction result; Generate a commodity preference recommendation table based on the commodity preference prediction result; Push the commodity data in the commodity preference recommendation table to the user irregularly in a preset order to form a personalized recommendation.

[0014] In a second aspect, the present invention provides a user portrait characterization system based on big data, which is characterized by including: An information graph construction module, configured to obtain user entities, commodity entities, and label entities based on a big data platform, and use the user entities, the commodity entities, and the label entities as nodes to construct a heterogeneous information graph, where the commodity entities are used to represent interaction objects with user entities, and the label entities are used to represent descriptive keywords defined by user entities and the big data platform for commodity entities; An attribute information acquisition module, configured to obtain initial user attribute information of a user entity based on the node connection relationship between the user entity and the commodity entity by using the heterogeneous information graph, and obtain label sentiment attribute information of a label entity based on the node connection relationship between the commodity entity and the label entity; A propagation matrix establishment module, configured to establish an attribute propagation matrix according to the node connection relationship between each node in the heterogeneous information graph; An attribute propagation iteration module, configured to input the initial user attribute information into a graph convolutional network, perform multiple rounds of attribute propagation based on the attribute propagation matrix, and iterate on the propagation weights between each node in each round of propagation to obtain user attribute information; A user portrait characterization module, configured to complete user portrait characterization by using the user attribute information, and apply the user portrait to personalized recommendation.

[0015] In a third aspect, the present invention provides an electronic device, 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 big data-based user portrait characterization method according to the first aspect or any possible design of the first aspect.

[0016] In a fourth aspect, the present invention provides a computer-readable storage medium, on which instructions are stored, and when the instructions run on a computer, the big data-based user portrait characterization method according to the first aspect or any possible design of the first aspect is executed.

[0017] In a fifth aspect, the present invention provides a computer program product containing instructions, and when the instructions run on a computer, the computer is made to execute the big data-based user portrait characterization method according to the first aspect or any possible design of the first aspect.

[0018] Beneficial effects: The present invention provides a method for depicting user portraits based on big data, which includes: First, based on a big data platform, user entities, commodity entities, and label entities are obtained, and a heterogeneous information graph is constructed using the user entities, commodity entities, and label entities as nodes. Among them, the commodity entity is used to represent the interaction object with the user entity, and the label entity is used to represent the descriptive keywords defined by the user entity and the big data platform for the commodity entity; Second, based on the heterogeneous information graph, using the node connection relationship between the user entity and the commodity entity, the initial user attribute information of the user entity is obtained, and using the node connection relationship between the commodity entity and the label entity, the label sentiment attribute information of the label entity is obtained; Then, according to the node connection relationship between each node in the heterogeneous information graph, an attribute propagation matrix is established; Then, the initial user attribute information is input into a graph convolutional network, and based on the attribute propagation matrix, multiple rounds of attribute propagation are performed, and the propagation weights between each node in each round of propagation are iterated to obtain user attribute information; Finally, using the user attribute information, the user portrait is depicted, and the user portrait is applied to personalized recommendation. By obtaining the label entity, the emotional characteristics of the user portrait are analyzed, and according to the node connection relationship between the user entity and the commodity entity, the preference characteristics of the user portrait for various commodities are analyzed to complete the deep characterization of the user portrait, and through multiple rounds of attribute propagation, accurate user attribute information is obtained to improve the accuracy of the user portrait. Description of the Drawings

[0019] Figure 1 It is a schematic flowchart of the steps of the method for depicting user portraits based on big data provided by an embodiment of the present invention; Figure 2 It is a schematic functional structure diagram of the system for depicting user portraits based on big data provided by an embodiment of the present invention; Figure 3 It is a schematic structural diagram of an electronic device provided by an embodiment of the present invention. Detailed Embodiments

[0020] 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 description of the embodiments or the prior art. Obviously, the following description of the structure of the drawings is 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 description of these embodiment modes is used to help understand the present invention, but does not constitute a limitation to the present invention.

[0021] 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 may be referred to as the second unit, and similarly, the second unit may be referred to as the first unit, without departing from the scope of the exemplary embodiments of the present invention.

[0022] It should be understood that for the term "and / or" that may appear herein, it is merely an association relationship describing associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, B exists alone, and both A and B exist simultaneously; for the term " / and" that may appear herein, it is a description of another associated object relationship, indicating that there can be two relationships. For example, A / and B can represent: A exists alone, and both A and B exist; additionally, for the character " / " that may appear herein, it generally represents that the associated objects before and after are in an "or" relationship.

[0023] Embodiment: As Figure 1 shown, in a first aspect of this embodiment, a method for depicting a user portrait based on big data is provided, which includes: S1. Based on a big data platform, obtain user entities, commodity entities, and label entities, and use the user entities, the commodity entities, and the label entities as nodes to construct a heterogeneous information graph, where the commodity entity is used to represent the interaction object with the user entity, and the label entity is used to represent the descriptive keywords defined by the user entity and the big data platform for the commodity entity; In a possible implementation manner, in step S1, based on the big data platform, obtaining user entities, commodity entities, and label entities can be decomposed into the following steps S11 - S15, including: S11. Obtain the user registration information filled in when the user registers, and based on the big data platform, obtain the user interaction behavior information of the user on the big data platform, and use the user registration information and the user interaction behavior information to form a user entity; S12. Based on the big data platform, obtain the metadata information of the commodity, and use the metadata information of the commodity as the commodity entity; S13. Based on the big data platform, obtain the evaluation information labeled by the user for the commodity, and combine the definition description of the commodity by the big data platform to form a label entity; Correspondingly, using the user entity, the commodity entity, and the label entity as nodes to construct a heterogeneous information graph includes: S14. Define the user entity as a user node, the commodity entity as a commodity node, and the label entity as a label node. Based on the big data platform, obtain the node connection relationships among the three types of nodes. Among them, an edge connection is formed between the user node and the commodity node through the user evaluation relationship, an edge connection is formed between the commodity node and the label node through the defined description relationship, and an edge connection is formed between the user node and the label node through the keyword usage relationship; S15. Based on the three types of nodes and the node connection relationships among the three types of nodes, construct a heterogeneous information graph.

[0024] It should be noted that the method of this embodiment can obtain the evaluations of users on various commodities by obtaining the label entity, and based on these evaluations, obtain the corresponding user emotional characteristics as the deep characteristics of the user entity, so that the depicted user portrait can have emotional characteristics to help complete the deep depiction of the user portrait.

[0025] S2. Based on the heterogeneous information graph, utilize the node connection relationship between the user entity and the commodity entity to obtain the initial user attribute information of the user entity, and utilize the node connection relationship between the commodity entity and the label entity to obtain the label emotional attribute information of the label entity; In a possible implementation manner, in step S2, based on the heterogeneous information graph, utilizing the node connection relationship between the user entity and the commodity entity to obtain the initial user attribute information of the user entity can be decomposed into the following steps S21 - S26, including: S21. Extract the user basic characteristics in the user entity, where the user basic characteristics include user age, user gender, user occupation, and user residence; S22. Based on the node connection relationship between the user node and the commodity node, calculate the commodity preference vectors of the user for different commodities respectively; S23. Combine the user basic characteristics and the commodity preference vectors of the user for different commodities to obtain the initial user attribute information of the user entity; Correspondingly, utilizing the node connection relationship between the commodity entity and the label entity to obtain the label emotional attribute information of the label entity includes: S24. Extract the evaluation behavior characteristics in the label entity, where the evaluation behavior characteristics include commodity score, evaluation type, and the proportion of each type of evaluation; S25. Based on the node connection relationship between the commodity entity and the label entity, perform emotional annotation on the keywords in the evaluation information labeled by the user for the commodity to extract the emotional preference vectors of the user for different labels; S26. Combine the evaluation behavior features and the user's emotional preference vectors for different tags to obtain the tag emotional attribute information of the tag entity.

[0026] It should be noted that the method of this embodiment can analyze the preference features of the user portrait for various commodities according to the node connection relationship between the user entity and the commodity entity. Combining the user emotional features, the user portrait can be completely depicted, making the completed user portrait more relevant to the commodities and tags, and being more capable of providing corresponding accurate commodity recommendations for the user portrait.

[0027] In a possible implementation manner, in step S25, the keyword sentiment annotation of the evaluation information annotated by the user for the commodity can be decomposed into the following steps S251 - S253, including: S251. Obtain the tag keywords in the tag entity to perform semantic disambiguation on the tag keywords to form unique tag keywords; S252. Obtain the NRC sentiment dictionary, and map the unique tag keywords to the NRC sentiment dictionary through the semantic similarity algorithm; If there is a matching sentiment category between the unique tag keyword and the NRC sentiment dictionary, count the occurrence frequency of the sentiment category, and select the sentiment category with the highest frequency as the sentiment-related word of the tag entity; If there is no matching sentiment category between the unique tag keyword and the NRC sentiment dictionary, for the unique tag keyword, use the semantic similarity threshold screening method combined with the context information to generate a sentiment-related word; S253. Use the sentiment-related word to perform sentiment annotation on the keywords in the evaluation information annotated by the user for the commodity.

[0028] S3. Establish an attribute propagation matrix according to the node connection relationship between each node in the heterogeneous information graph; In a possible implementation manner, in step S3, establishing an attribute propagation matrix according to the node connection relationship between each node in the heterogeneous information graph can be decomposed into the following steps S31 - S32, including: S31. According to the node connection relationship between each node in the heterogeneous information graph, assign initial propagation weights to each node connection relationship according to the node connection relationship among the three types of nodes. Among them, the initial propagation weights include the first edge propagation weight, the second edge propagation weight, and the third edge propagation weight. And the first edge propagation weight is the user's evaluation of the commodity, the second edge propagation weight is the description frequency of the tag for the commodity, and the third edge propagation weight is the frequency of the user using the tag; S32. Based on the number of nodes in the heterogeneous information graph, establish a weight matrix, fill the initial propagation weights into the weight matrix, and perform normalization processing to use the weight matrix as the attribute propagation matrix.

[0029] S4. Input the user's initial attribute information into the graph convolutional network, and based on the attribute propagation matrix, perform multiple rounds of attribute propagation and iterate the propagation weights between nodes in each round to obtain the user's attribute information. In a possible implementation manner, in step S4, inputting the user's initial attribute information into the graph convolutional network and performing multiple rounds of attribute propagation based on the attribute propagation matrix can be, but is not limited to, decomposed into the following steps S41 - S44, including: S41. Obtain the metadata information of the commodity, the user's initial attribute information, and the label sentiment attribute information as the initial attribute propagation information, and input the initial attribute propagation information as the input quantity into the graph convolutional network. S42. For each node corresponding to a user entity, use the attribute propagation matrix to perform weighted aggregation on the attribute information of adjacent nodes to perform multiple rounds of attribute propagation. Correspondingly, iterating the propagation weights between nodes in each round to obtain the user's attribute information includes: S43. In multiple rounds of attribute propagation, fuse the attribute propagation information in each round of attribute propagation with the initial attribute propagation information through a smoothing factor to complete the update of the attribute propagation information. S44. Obtain the preset maximum number of iterations and the preset change rate threshold. When the change rate of the attribute propagation information is lower than the preset change rate threshold or reaches the maximum number of iterations, stop the attribute propagation to obtain the final attribute propagation information, and use the final attribute propagation information as the user's attribute information, where the user's attribute information includes commodity preference, sentiment preference, and evaluation behavior.

[0030] S5. Use the user's attribute information to complete the user portrait description and apply the user portrait to personalized recommendation.

[0031] In a possible implementation manner, in step S5, using the user's attribute information to complete the user portrait description and apply the user portrait to personalized recommendation can be, but is not limited to, decomposed into the following steps S51 - S53, including: S51. Obtain the personalized prediction model, input the user portrait into the commodity preference prediction model to output the commodity preference prediction result. S52. Generate a commodity preference recommendation table based on the commodity preference prediction result. S53. Push the commodity data in the commodity preference recommendation table to the user irregularly in a preset order to form a personalized recommendation.

[0032] As Figure 2 shown, the second aspect of this embodiment provides a hardware system for implementing the big data-based user portrait characterization method described in the first aspect of the embodiment, including: An information graph construction module, configured to obtain user entities, commodity entities, and tag entities based on a big data platform, and use the user entities, the commodity entities, and the tag entities as nodes to construct a heterogeneous information graph, where the commodity entity is used to represent an interaction object with the user entity, and the tag entity is used to represent descriptive keywords defined by the user entity and the big data platform for the commodity entity; An attribute information acquisition module, configured to obtain initial user attribute information of the user entity based on the node connection relationship between the user entity and the commodity entity in the heterogeneous information graph, and obtain tag sentiment attribute information of the tag entity based on the node connection relationship between the commodity entity and the tag entity; A propagation matrix establishment module, configured to establish an attribute propagation matrix according to the node connection relationship between each node in the heterogeneous information graph; An attribute propagation iteration module, configured to input the initial user attribute information into a graph convolutional network, perform multiple rounds of attribute propagation based on the attribute propagation matrix, and iterate the propagation weights between each node in each round of propagation to obtain user attribute information; A user portrait characterization module, configured to complete user portrait characterization by using the user attribute information, and apply the user portrait to personalized recommendation.

[0033] For the working process, working details, and technical effects of the system provided in this embodiment, reference may be made to the first aspect of the embodiment, which will not be elaborated here.

[0034] As Figure 3 shown, the third aspect of this embodiment provides an electronic device, 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 big data-based user portrait characterization method described in the first aspect of the embodiment.

[0035] Specifically, the memory may include, but is not limited to, random access memory (RAM), read only memory (ROM), flash memory, first input first output (FIFO), and / or first in last out (FILO), etc.; specifically, the processor may include one or more processing cores, such as a 4-core processor, an 8-core processor, etc. The processor may be implemented in at least one hardware form of DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), or PLA (Programmable Logic Array). At the same time, the processor may also include a main processor and a co-processor. The main processor is a processor used to process data in the wake state, also known as the CPU (Central Processing Unit); the co-processor is a low-power processor used to process data in the standby state.

[0036] In some embodiments, the processor may be integrated with a GPU (Graphics Processing Unit), and the GPU is responsible for rendering and drawing the content to be displayed on the display screen. For example, the processor may not be limited to using a microprocessor of the STM32F105 series, a reduced instruction set computer (RISC) microprocessor, a processor with an X86 architecture, or a processor integrated with a neural-network processing unit (NPU); the transceiver may include, but is not limited to, a Wi-Fi wireless transceiver, a Bluetooth wireless transceiver, a General Packet Radio Service (GPRS) wireless transceiver, a ZigBee wireless transceiver (a low-power local area network protocol based on the IEEE802.15.4 standard), a 3G transceiver, a 4G transceiver, and / or a 5G transceiver, etc. In addition, the device may also 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 electronic device provided in this embodiment, reference may be made to the first aspect of the embodiment, which will not be elaborated here.

[0038] The fourth aspect of this embodiment provides a storage medium storing instructions for the big data-based user portrait characterization method described in the first aspect of the embodiment, that is, instructions are stored on the storage medium, and when the instructions run on a computer, they execute the big data-based user portrait characterization method described in the first aspect of the embodiment.

[0039] Among them, the storage medium refers to a carrier for storing data, and may include, but is not limited to, floppy disks, optical discs, hard disks, flash memories, USB flash drives, and / or Memory Sticks, etc. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices.

[0040] For the working process, working details, and technical effects of the storage medium provided in this embodiment, reference may be made to the first aspect of the embodiment, which will not be elaborated here.

[0041] The fifth aspect of this embodiment provides a computer program product containing instructions, which, when running on a computer, cause the computer to execute the big data-based user portrait characterization method described in the first aspect of the embodiment, where the computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices.

[0042] Finally, it should be noted that the above are only preferred embodiments of the present invention and are not used 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 method for depicting user portraits based on big data, characterized in that, Including: Based on a big data platform, user entities, commodity entities, and label entities are obtained, and a heterogeneous information graph is constructed using the user entities, the commodity entities, and the label entities as nodes. Among them, the commodity entities are used to represent interaction objects with user entities, and the label entities are used to represent descriptive keywords defined by user entities and the big data platform for commodity entities; Based on the heterogeneous information graph, using the node connection relationship between the user entity and the commodity entity, the initial user attribute information of the user entity is obtained, and using the node connection relationship between the commodity entity and the label entity, the label sentiment attribute information of the label entity is obtained; According to the node connection relationship between each node in the heterogeneous information graph, an attribute propagation matrix is established; The initial user attribute information is input into a graph convolutional network, and based on the attribute propagation matrix, multiple rounds of attribute propagation are performed, and the propagation weights between each node in each round of propagation are iterated to obtain user attribute information; Using the user attribute information, user portrait characterization is completed to apply the user portrait to personalized recommendation.

2. The method for depicting user portraits based on big data according to claim 1, wherein Based on a big data platform, obtaining user entities, commodity entities, and label entities includes: Obtaining user registration information filled in by users during registration, and based on the big data platform, obtaining user interaction behavior information of users on the big data platform, so as to form user entities using the user registration information and the user interaction behavior information; Based on the big data platform, obtaining the metadata information of commodities, and using the metadata information of commodities as commodity entities; Based on the big data platform, obtaining evaluation information labeled by users for commodities, and combining the definition descriptions of commodities by the big data platform to form label entities; Correspondingly, using the user entities, the commodity entities, and the label entities as nodes to construct a heterogeneous information graph includes: Defining the user entity as a user node, the commodity entity as a commodity node, and the label entity as a label node, and obtaining the node connection relationship between the three types of nodes according to the big data platform. Among them, an edge connection is formed between the user node and the commodity node through a user evaluation relationship, an edge connection is formed between the commodity node and the label node through a definition description relationship, and an edge connection is formed between the user node and the label node through a keyword usage relationship; Based on the three types of nodes and the node connection relationship between the three types of nodes, a heterogeneous information graph is constructed.

3. The method for depicting user portraits based on big data according to claim 1, wherein Based on the heterogeneous information graph, using the node connection relationship between the user entity and the commodity entity to obtain the initial user attribute information of the user entity includes: Extracting the basic user features in the user entity, where the basic user features include user age, user gender, user occupation, and user residence; Based on the node connection relationship between the user entity and the commodity entity, the commodity preference vectors of users for different commodities are calculated respectively; Combining the basic user features and the commodity preference vectors of users for different commodities to obtain the initial user attribute information of the user entity; Correspondingly, using the node connection relationship between the commodity entity and the label entity, the label sentiment attribute information of the label entity is obtained, including: Extract the evaluation behavior characteristics in the label entity, where the evaluation behavior characteristics include commodity ratings, evaluation types, and the proportion of each type of evaluation; Based on the node connection relationship between the commodity entity and the label entity, perform sentiment annotation on the keywords in the evaluation information annotated by the user for the commodity, so as to extract the sentiment preference vectors of the user for different labels; Combine the evaluation behavior characteristics and the sentiment preference vectors of the user for different labels to obtain the label sentiment attribute information of the label entity.

4. The method for depicting user portraits based on big data according to claim 3, wherein Performing sentiment annotation on the keywords in the evaluation information annotated by the user for the commodity includes: Obtain the label keywords in the label entity to perform semantic disambiguation on the label keywords to form unique label keywords; Obtain the NRC sentiment dictionary, and map the unique label keywords to the NRC sentiment dictionary through a semantic similarity algorithm; If there is a matching sentiment category between the unique label keyword and the NRC sentiment dictionary, count the occurrence frequency of the sentiment category, and select the sentiment category with the highest frequency of the sentiment category as the sentiment-related word of the label entity; If there is no matching sentiment category between the unique label keyword and the NRC sentiment dictionary, for the unique label keyword, use a semantic similarity threshold screening method combined with context information to generate a sentiment-related word; Use the sentiment-related word to perform sentiment annotation on the keywords in the evaluation information annotated by the user for the commodity.

5. The method for depicting user portraits based on big data according to claim 1, wherein According to the node connection relationship between each node in the heterogeneous information graph, an attribute propagation matrix is established, including: According to the node connection relationship between each node in the heterogeneous information graph, and according to the node connection relationship between the three types of nodes, assign initial propagation weights to each node connection relationship, where the initial propagation weights include the first edge propagation weight, the second edge propagation weight, and the third edge propagation weight, and the first edge propagation weight is the user's evaluation of the commodity, the second edge propagation weight is the description frequency of the label for the commodity, and the third edge propagation weight is the frequency of the user using the label; Based on the number of nodes in the heterogeneous information graph, establish a weight matrix, fill the initial propagation weights into the weight matrix, and perform normalization processing to use the weight matrix as the attribute propagation matrix.

6. The method for depicting user portraits based on big data according to claim 1, wherein Input the user's initial attribute information into the graph convolutional network, and based on the attribute propagation matrix, perform multiple rounds of attribute propagation, including: Obtain the metadata information of the commodity, the user's initial attribute information, and the label sentiment attribute information as the initial attribute propagation information, and input the initial attribute propagation information as the input quantity into the graph convolutional network; For each node corresponding to the user entity, use the attribute propagation matrix to perform weighted aggregation on the attribute information of adjacent nodes to perform multiple rounds of attribute propagation; Correspondingly, iterate the propagation weights between each node in each round of propagation to obtain the user attribute information, including: In multiple rounds of attribute propagation, the attribute propagation information in each round of attribute propagation is fused with the initial attribute propagation information through a smoothing factor to complete the update of the attribute propagation information; Obtain a preset maximum number of iterations and a preset change rate threshold. When the change rate of the attribute propagation information is lower than the preset change rate threshold or reaches the maximum number of iterations, stop the attribute propagation to obtain the final attribute propagation information, and use the final attribute propagation information as user attribute information, where the user attribute information includes commodity preferences, emotional preferences, and evaluation behaviors.

7. The method for depicting user portraits based on big data according to claim 1, wherein Using the user attribute information, complete the user portrait characterization to apply the user portrait to personalized recommendation, including: Obtain a personalized prediction model, and input the user portrait into the commodity preference prediction model to output a commodity preference prediction result; Based on the commodity preference prediction result, generate a commodity preference recommendation table; According to the preset order, push the commodity data in the commodity preference recommendation table to the user irregularly to form a personalized recommendation.

8. A user portrait characterization system based on big data, characterized in that, Including: An information graph construction module, configured to obtain user entities, commodity entities, and tag entities based on a big data platform, and use the user entities, the commodity entities, and the tag entities as nodes to construct a heterogeneous information graph, where the commodity entity is used to represent the interaction object with the user entity, and the tag entity is used to represent the descriptive keywords defined by the user entity and the big data platform for the commodity entity; An attribute information acquisition module, configured to obtain the initial user attribute information of the user entity based on the node connection relationship between the user entity and the commodity entity in the heterogeneous information graph, and obtain the tag emotional attribute information of the tag entity based on the node connection relationship between the commodity entity and the tag entity; A propagation matrix establishment module, configured to establish an attribute propagation matrix according to the node connection relationship between each node in the heterogeneous information graph; An attribute propagation iteration module, configured to input the initial user attribute information into a graph convolutional network, perform multiple rounds of attribute propagation based on the attribute propagation matrix, and iterate the propagation weights between each node in each round of propagation to obtain user attribute information; A user portrait characterization module, configured to use the user attribute information to complete the user portrait characterization to apply the user portrait to personalized recommendation.

9. An electronic 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 big data-based user portrait characterization method according to any one of claims 1 to 7.

10. A computer program product comprising a computer program or instructions, characterized in that, The computer program or the instruction, when executed by a computer, implements the big data-based user portrait characterization method according to any one of claims 1 to 7.

Images