User solution recommendation method based on enterprise user portrait analysis

Abstract

The invention discloses a user solution recommendation method based on enterprise user portrait analysis. The recommendation method comprises the following steps: S1, collecting first enterprise information of a to-be-recommended target user and second enterprise information of a historical user forming a solution; s2, forming a to-be-recommended user portrait and a historical user portrait based on the first enterprise information and the second enterprise information; s3, forming a hypergraph network by taking a plurality of historical user portrait vectorization as object nodes and taking a plurality of enterprise problem solution vectorization as result nodes; s4, performing community division on the hypergraph network; s5, searching a community most related to the user problem vector from the hypergraph network; and S6, performing similarity comparison on the object node in the community most related to the user problem vector and the to-be-recommended user portrait, selecting the object node with high similarity as the closest enterprise, and taking the enterprise problem solution corresponding to the object node as a solution recommendation result.

Description

Technical Field

[0001] The present invention relates to the technical field of enterprise solution recommendation, and in particular to a method for recommending user solutions based on enterprise user portrait analysis. Background Art

[0002] Enterprise management consulting is also known as "enterprise diagnosis". Enterprise management experts use scientific methods to find out the problems in enterprise management based on investigation and analysis, propose specific improvement plans, and guide their implementation to help enterprises improve their business management. It is a service activity. Its characteristics are: (1) service; (2) guidance; (3) scientific; (4) practical; (5) arduous. In the process of enterprise management consulting service management in the existing technology, manual customer service is often used to connect with relevant consulting experts to solve the business management difficulties that arise in the process of enterprise management. Enterprise managers readjust enterprise management based on the answers provided by consulting experts. However, consulting experts are usually not more familiar with the enterprise itself than enterprise managers. In addition, current enterprise managers hope to obtain enterprise management solutions that can be referenced by other enterprises, share the management resources of various enterprises, especially some good management resources, so that other enterprises of the same type can learn from enterprise management resources to improve the management efficiency of enterprises. However, there is currently a lack of enterprise solution recommendation methods that can be based on user needs. Summary of the invention

[0003] In view of the above-mentioned prior art, the present invention provides a user solution recommendation method based on enterprise user portrait analysis, mainly to solve the technical problems existing in the above-mentioned background technology.

[0004] To achieve the above object, the technical solution of the embodiment of the present invention is implemented as follows:

[0005] A user solution recommendation method based on enterprise user portrait analysis, the recommendation method comprising the following steps:

[0006] S1. Determine the target user to be recommended, collect the first enterprise information of the target user to be recommended and the second enterprise information of the historical user who has formed a solution, and collect the enterprise problem solutions corresponding to the historical user;

[0007] S2. Forming a to-be-recommended user portrait and a historical user portrait based on the first enterprise information and the second enterprise information;

[0008] S3, vectorizing multiple historical user portraits as object nodes, vectorizing multiple enterprise problem solutions as result nodes, establishing a first edge based on the object nodes and the result nodes, and constructing a second edge based on multiple result nodes, wherein the set of the first edge, the second edge, the object node, and the result node constitute a hypergraph network;

[0009] S4, performing community division on the hypergraph network;

[0010] S5. Obtain the question input by the user and convert it into a user question vector, and search the community most relevant to the user question vector from the hypergraph network;

[0011] S6. Compare the similarity between the object node in the community most related to the user problem vector and the user portrait to be recommended, select the object node with high similarity as the closest enterprise, and its corresponding enterprise problem solution as the solution recommendation result.

[0012] Optionally, information data of target users to be recommended and historical users for whom solutions have been formed are collected through database export, web crawler selection, and manual input.

[0013] Optionally, forming a to-be-recommended user portrait and a historical user portrait based on the first enterprise information and the second enterprise information specifically includes:

[0014] Constructing an enterprise portrait model, which consists of five dimensions: basic attributes of the enterprise, business scope of the enterprise, main products of the enterprise, internal and external connections of the enterprise, and industry evaluation of the enterprise;

[0015] Extract keywords from the first enterprise information as feature items, and fill the feature items into five corresponding dimensions of the enterprise's basic attributes, enterprise business scope, enterprise's main products, enterprise's internal and external connections, and enterprise's industry evaluation in the enterprise portrait model to form a user portrait to be recommended;

[0016] The keywords in the second enterprise information are extracted as feature items, and the feature items are filled into the five corresponding dimensions of the basic attributes of the enterprise, the business scope of the enterprise, the main products of the enterprise, the internal and external connections of the enterprise, and the industry evaluation of the enterprise in the new enterprise portrait model to form a historical user portrait.

[0017] Optionally, step S3 specifically includes:

[0018] Extract key sentences from the enterprise problem solution, and vectorize the key sentences and the enterprise problem solution as result nodes;

[0019] Based on the correspondence between historical users and enterprise problem solutions, a first edge is established between an object node and a result node where the correspondence exists;

[0020] Calculate the similarity between two key sentences in all pairwise result nodes. If the similarity between any two key sentences is greater than the first threshold, establish a second edge between the two result nodes, and the value of the similarity is the information of the second edge.

[0021] A hypergraph network is constructed according to all the first edges and the second edges, and the hypergraph network includes result nodes and object nodes.

[0022] Optionally, the Louvain algorithm is used to divide the hypergraph network into communities, and the divided communities all contain a certain number of first edges, second edges, result nodes, and object nodes.

[0023] Optionally, obtaining the question input by the user and converting it into a user question vector, and searching the hypergraph network for a community most relevant to the user question vector, specifically includes:

[0024] Use vector tools to vectorize the questions input by users, and convert the questions input by users into user question vectors;

[0025] The similarity between the user question vector and all result nodes in each community in the hypergraph network is compared and calculated in turn to find the community most relevant to the user question vector.

[0026] Optionally, in any community, if the number of result nodes that meet the similarity requirement with the user question vector is greater than a second threshold, it means that the community is the community most relevant to the user question vector.

[0027] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, a user solution recommendation method based on enterprise user portrait analysis as described in any of the foregoing items is implemented.

[0028] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the method for recommending user solutions based on enterprise user portrait analysis as described in any of the above items is implemented.

[0029] The beneficial effects of the present invention are as follows: the user solution recommendation method based on enterprise user portrait analysis provided by the present application, by constructing a hypergraph network, vectorizes historical user portraits and solutions into object nodes and result nodes, and establishes edges based on the relationship between them, forming a complex network structure. This structure not only contains the direct connection between enterprise information and solutions, but also reveals the potential associations between different solutions, providing richer information support for the recommendation system. The construction of the hypergraph network enables the recommendation system to consider various factors from a global perspective, thereby providing a more reasonable and effective solution. By vectorizing the problem input by the user and comparing the similarity with the community in the hypergraph network, the most relevant community can be quickly located. This process not only improves the search efficiency, but also ensures the relevance and accuracy of the recommendation results. By comparing the similarity of the object nodes in the most relevant community with the user portrait to be recommended, the object node with the highest similarity is selected as the closest enterprise, and its corresponding enterprise problem solution is used as the final recommendation result, realizing a highly personalized recommendation service. This method not only takes into account the current problems faced by users, but also fully combines the user's portrait characteristics to realize personalized recommendations, which helps to improve user satisfaction and user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 Schematic diagram of the process of recommending user solutions based on enterprise user portrait analysis in an embodiment of the present application. DETAILED DESCRIPTION

[0031] The technical solution of the present invention is further elaborated in detail below in conjunction with the drawings and specific embodiments of the specification. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by technicians in the technical field of the present invention. The terms used in the specification of the present invention herein are only for the purpose of describing specific embodiments and are not intended to limit the present invention. In the following description, the expression "some embodiments" is related to a subset of all possible embodiments, but it should be understood that "some embodiments" can be the same subset or different subsets of all possible embodiments, and can be combined with each other without conflict.

[0032] In the following description, a large number of specific details are provided to provide a more thorough understanding of the present invention. However, it is apparent to those skilled in the art that the present invention can be implemented without one or more of these details. In other examples, in order to avoid confusion with the present invention, some technical features well known in the art are not described.

[0033] It should be understood that the present invention can be implemented in different forms and should not be interpreted as being limited to the embodiments proposed herein. On the contrary, providing these embodiments will make the disclosure thorough and complete, and the scope of the present invention will be fully conveyed to those skilled in the art. And the purpose of the terms used herein is only to describe specific embodiments and is not intended to be a limitation of the present invention. When used herein, the singular forms of "one", "one" and "said / the" are also intended to include plural forms, unless the context clearly indicates another way. It should also be understood that the terms "compose" and / or "include" when used in this specification determine the presence of the features, integers, steps, operations, elements and / or parts, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, parts and / or groups. When used herein, the term "and / or" includes any and all combinations of the relevant listed items.

[0034] It should also be noted that when an element is referred to as being "fixed to" another element, it may be directly on the other element or there may be a central element. When an element is considered to be "connected to" another element, it may be directly connected to the other element or there may be a central element at the same time. The terms "vertical", "horizontal", "inside", "outside", "left", "right" and similar expressions used herein are for illustrative purposes only and are not intended to be the only implementation method.

[0035] In order to fully understand the present invention, a detailed structure will be proposed in the following description to illustrate the technical solution proposed by the present invention. The optional embodiments of the present invention are described in detail as follows, but in addition to these detailed descriptions, the present invention may also have other implementations.

[0036] Please refer to the attached Figure 1 , the present application provides a user solution recommendation method based on enterprise user portrait analysis, the recommendation method comprising the following steps:

[0037] S1. Determine the target user to be recommended, collect the first enterprise information of the target user to be recommended and the second enterprise information of the historical user who has formed a solution, and collect the enterprise problem solutions corresponding to the historical user;

[0038] S2. Forming a to-be-recommended user portrait and a historical user portrait based on the first enterprise information and the second enterprise information;

[0039] S3, vectorizing multiple historical user portraits as object nodes, vectorizing multiple enterprise problem solutions as result nodes, establishing a first edge based on the object nodes and the result nodes, and constructing a second edge based on multiple result nodes, wherein the set of the first edge, the second edge, the object node, and the result node constitute a hypergraph network;

[0040] S4, performing community division on the hypergraph network;

[0041] S5. Obtain the question input by the user and convert it into a user question vector, and search the community most relevant to the user question vector from the hypergraph network;

[0042] S6. Compare the similarity between the object node in the community most related to the user problem vector and the user portrait to be recommended, select the object node with high similarity as the closest enterprise, and its corresponding enterprise problem solution as the solution recommendation result.

[0043] In step S1, the target user is determined in order to clarify the object to which the solution needs to be recommended, which is usually a corporate user, to ensure that the subsequent data collection and analysis are targeted. The first corporate information collected from the target user includes the basic information of the company (such as name, date of establishment, registered capital, legal representative, corporate address, etc.), business scope, main products, internal and external connections (such as equity structure, partners, etc.) and industry evaluation (such as honors obtained, market position, etc.). This information can be obtained through database export, web crawler selection, manual input, etc. This information will be used to build a portrait of the target user for subsequent similarity calculation and recommendation.

[0044] At the same time, the second enterprise information of the historical users who have formed the solution is collected. This information is similar to the first enterprise information of the target user, including the basic information, business scope, main products, internal and external connections and industry evaluation of the historical users. This information can also be obtained through database export, web crawler selection, manual input, etc. The enterprise information of the historical users will be used to build the portrait of the historical users for subsequent similarity calculation and recommendation. The historical users are usually also enterprise users.

[0045] In addition, it is also necessary to collect the enterprise problem solutions corresponding to historical users. These solutions include various problems encountered by historical users in the past and their corresponding solutions. Problems can be technical problems, management problems, market problems, etc., and solutions can be specific operation steps, strategy suggestions, tool recommendations, etc. These solutions can be collected through historical records, customer feedback, technical support documents, etc. These solutions will be used to build result nodes in the hypergraph network to provide data support for subsequent recommendations.

[0046] This step can improve the accuracy of recommendations. Building a detailed corporate portrait can include the multi-dimensional characteristics of the company, which is helpful for subsequent similarity calculation and recommendation. Collecting a large number of historical user problem solutions can provide rich data support for the recommendation system, enabling it to provide more accurate and effective solutions.

[0047] In addition, the collection channels of enterprise information data include one or more of the following: the National Enterprise Credit Information Publicity System Database, the official registered enterprise credit reporting agency database under the National Small and Medium Enterprise Development Fund, the database of relevant cooperative units, the official website database of enterprises, the Tianyancha database, and the Qichacha database. The collection methods include database export, crawler crawling, and manual collection and input;

[0048] The information data of the target enterprise are classified into basic attribute data of the enterprise, business scope data of the enterprise, main product data of the enterprise, internal and external correlation data of the enterprise, and industry evaluation data of the enterprise.

[0049] Among them, the basic attribute data of the enterprise include the enterprise name, establishment date, registered capital, legal representative, and enterprise address; the business scope data of the enterprise are the commodity categories, varieties and service items produced and operated by the enterprise, reflecting the content of the enterprise's business activities and the production and operation direction; the main product data of the enterprise are the main products produced by the enterprise, and also include the representative products of the enterprise sold on the market, which is a further specific description of the enterprise's production; the internal and external related data of the enterprise include the enterprise's shareholding holding company and the enterprise's held holding company; the enterprise industry evaluation data include the relevant awards won by the enterprise.

[0050] The pre-processing methods for the collected enterprise information data include: data cleaning, data aggregation, and data deletion of the target enterprise's information data;

[0051] The information data after data deletion is converted from text data to structured data. When converting the information data from text data to structured data, it is necessary to identify the named entities in the text. Named entities mainly include name entities (company names, personal names, place names), time expressions (dates, times) and numerical expressions (currency values, percentages, etc.). Among them, for the identification of company names, personal names and place names, named entity recognition is implemented through rule-based methods, statistical-based machine learning methods and deep learning-based methods, as well as calling the domestic open source Chinese language processing tool NLPIR to extract information from the text;

[0052] There are problems of missing data, data duplication, and homonymy in structured data. For the problem of missing data, the information can be supplemented or improved through search engines or by temporarily leaving it blank and supplementing it later; for the problem of duplicate data, it is necessary to delete redundant information to ensure the uniqueness of the data, and the remaining information complements each other; for the problem of homonymy, it is necessary to solve it through disambiguation, and the methods used for disambiguation include clustering-based disambiguation and entity link-based disambiguation. At the same time, in order to solve the problem that the metadata standards used for information from different sources in the original data are different, a unified description of enterprise information is made, and data integration is carried out to integrate data from multiple data sources for unified storage. The above methods can eliminate the problems of missing data, data duplication, and homonymy in structured data.

[0053] In an optional implementation, forming a to-be-recommended user portrait and a historical user portrait based on the first enterprise information and the second enterprise information specifically includes:

[0054] Constructing an enterprise portrait model, which consists of five dimensions: basic attributes of the enterprise, business scope of the enterprise, main products of the enterprise, internal and external connections of the enterprise, and industry evaluation of the enterprise;

[0055] Extract keywords from the first enterprise information as feature items, and fill the feature items into five corresponding dimensions of the enterprise's basic attributes, enterprise business scope, enterprise's main products, enterprise's internal and external connections, and enterprise's industry evaluation in the enterprise portrait model to form a user portrait to be recommended;

[0056] The keywords in the second enterprise information are extracted as feature items, and the feature items are filled into the five corresponding dimensions of the basic attributes of the enterprise, the business scope of the enterprise, the main products of the enterprise, the internal and external connections of the enterprise, and the industry evaluation of the enterprise in the new enterprise portrait model to form a historical user portrait.

[0057] Specifically, the enterprise portrait is a multi-dimensional, multi-level user model. According to the type of information data of the target enterprise, a vector space is defined as the enterprise portrait. The dimensional features that make up the enterprise portrait include basic attributes of the enterprise, business scope of the enterprise, main products of the enterprise, enterprise association analysis, and enterprise industry evaluation. The specific model structure is as follows:

[0058] Enterprise=<BaseInfo,BusiScope,MainProduct,Relation,Comments>

[0059] Among them, BaseInfo represents the basic attributes of the enterprise, BusiScope represents the business scope of the enterprise, MainProduct represents the main products of the enterprise, Relation represents the enterprise relationship analysis, and Comments represents the industry evaluation of the enterprise. The five-tuple feature labels together constitute the enterprise portrait.

[0060] Extract feature items from the preprocessed enterprise information data and fill them into the corresponding dimension features to obtain the following results:

[0061] BaseInfo=<Name,EstabliDate,RegistCapital,LegalRepre,OrgaAddr> , which represent the company name, establishment date, registered capital, legal representative and company address respectively. The basic attribute characteristics of the company provide the most basic understanding and description of the company. This feature is relatively clear and relatively standardized. The basic information is concise, and the information in the database can be directly used as the feature item;

[0062] BusiScope={(Top1,t1),(Top2,t2),(Top3,t3),…,(Top n ,t n )} Among them, Top n represents the nth business direction of the enterprise, t n Indicates the business relevance of the enterprise to the nth direction, t n The larger the value is, the higher the corresponding business relevance of the enterprise. The business scope characteristics of the enterprise use keywords as the main representation of the business scope. The extracted keywords are a highly condensed and summarized summary of the enterprise's business direction.

[0063] The process of extracting keywords includes: using TF-IDF to pre-process the text of the business scope in the enterprise information database to obtain a more reasonable vector that can better represent the characteristics of the document. After converting it into a document vector, keyword extraction is performed based on the weight size.

[0064] In addition, we conduct topic analysis on business keywords of enterprises, and the main method used is LDA document topic generation model to analyze the characteristics of business scope of enterprises.

[0065] MainProduct= <p1,p2,p3,…,p n >, where p represents the main representative product of the target company. The main product features of the enterprise directly adopt the main products of the enterprise and the names of the main commodities sold by the enterprise on the market as characteristic items, which enhances the public's understanding of the enterprise's product line to a certain extent.

[0066] Relation=<type,object,percent> Among them, type is the type of enterprise holding or being held, object is the object of enterprise holding or being held, and percent is the equity ratio. The enterprise association analysis feature uses the data in the enterprise information database, according to the type of the enterprise holding or being held, and according to the calculation method of equity penetration, it traverses and calculates the equity relationship between the enterprise and related enterprises one by one. This dimension feature uses the name of the relevant company as the feature item and the equity ratio as the weight to intuitively reflect the internal and external connections of the enterprise.

[0067] Comments= <c1,c2,c3,…,c n >, where c represents the relevant honorary awards won by the enterprise. The enterprise industry evaluation characteristics directly use the honorary awards won by the enterprise as feature items, and also serve as an external display of excellent enterprises.

[0068] It should also be noted that among the above five dimensional characteristics, the characteristic items in the basic attribute characteristics of the enterprise, the main product characteristics of the enterprise, and the industry evaluation characteristics of the enterprise do not have weight factors;

[0069] The characteristic items in the enterprise business scope characteristics and enterprise association analysis characteristics have weight factors, and the weight calculation of the characteristic items needs to be implemented based on the weights of TF-IDF and LDA feature analysis.

[0070] In an optional implementation, step S3 specifically includes:

[0071] Extract key sentences from the enterprise problem solution, and vectorize the key sentences and the enterprise problem solution as result nodes;

[0072] Based on the correspondence between historical users and enterprise problem solutions, a first edge is established between an object node and a result node where the correspondence exists;

[0073] Calculate the similarity between two key sentences in all pairwise result nodes, wherein similarity calculation methods include cosine similarity, Euclidean distance, Jaccard similarity coefficient, etc. This embodiment adopts the cosine similarity calculation method for calculation, and its specific formula is common knowledge of those skilled in the art, and is not specifically described in this embodiment. If the similarity between any two key sentences is greater than the first threshold, a second edge is established between the two result nodes, and the value of the similarity is the information of the second edge;

[0074] A hypergraph network is constructed according to all the first edges and the second edges, and the hypergraph network includes result nodes and object nodes.

[0075] Furthermore, in this embodiment, the first threshold is set to 0.8, and if the similarity between two vectors is higher than the threshold, an edge is established between the two nodes. However, it should be noted that those skilled in the art may also reset other thresholds according to actual conditions.

[0076] Furthermore, the Louvain algorithm is used to divide the hypergraph network into communities, and the divided communities contain a certain number of first edges, second edges, result nodes, and object nodes. The communities formed can be financial communities, marketing communities, personnel communities, etc., which include a variety of solutions related to the business operations of the enterprise.

[0077] Further, the question input by the user is obtained and converted into a user question vector. For example, the question input by the user is vectorized using a PQ model, the question input by the user is converted into a user question vector, and the community most relevant to the user question vector is searched from the hypergraph network, specifically including:

[0078] Use vector tools to vectorize the questions input by users, and convert the questions input by users into user question vectors;

[0079] The similarity between the user question vector and all the result nodes in each community in the hypergraph network is compared and calculated in turn to find the community most relevant to the user question vector. In any community, if the number of result nodes that meet the similarity requirement with the user question vector is greater than a second threshold, it means that the community is the community most relevant to the user question vector.

[0080] Exemplarily, the cosine similarity calculation formula is used to calculate the similarity between the user question vector and all result nodes in each community. For example, there are result nodes such as AF in community A. The similarity between the user question vector and result nodes such as AF is calculated respectively. If the number of result nodes with a similarity higher than 0.7 is greater than 5, it means that community A is the most relevant community.

[0081] Then, a similarity algorithm is used to compare the similarity of all object nodes in community A with the user portrait to be recommended, and the object node with high similarity is selected as the closest enterprise, and its corresponding enterprise problem solution is used as the solution recommendation result.

[0082] In summary, the present invention provides personalized solution recommendations for target users to be recommended by constructing an enterprise user portrait model and utilizing the data and solutions of historical users. First, by collecting, preprocessing and structurally converting the information data of target users and historical users, a detailed enterprise portrait is formed, ensuring that the data basis of the recommendation system is accurate and reliable. The collection of these information data is not limited to database export, but also includes a variety of methods such as web crawler selection and manual input, ensuring the comprehensiveness and timeliness of the information.

[0083] The construction of the enterprise portrait model is completed by defining five dimensional features: basic attributes of the enterprise, business scope of the enterprise, main products of the enterprise, internal and external connections of the enterprise, and industry evaluation of the enterprise. Each dimensional feature extracts keywords as feature items from the pre-processed enterprise information data through specific algorithms and technical means, and fills them into the corresponding dimensions to form target user portraits and historical user portraits. This multi-dimensional enterprise portrait model can comprehensively and accurately depict the characteristics of the enterprise and improve the accuracy of recommendations.

[0084] Next, the present invention constructs a hypergraph network by vectorizing historical user portraits into object nodes, vectorizing enterprise problem solutions as result nodes, establishing a first edge based on the relationship between the two, and establishing a second edge based on the similarity between the result nodes. The hypergraph network not only contains the direct connection between enterprise information and solutions, but also reflects the correlation between different solutions, providing structured data support for subsequent community division and similarity calculation.

[0085] The community division of the Hypergraph network uses the Louvain algorithm, which can efficiently discover the community structure in the network, allowing the recommendation system to quickly locate the community most relevant to the user's question in a large-scale data set. By converting the user's input question into a vector form and comparing the similarity with each community in the Hypergraph network, the most relevant community can be accurately found, thereby improving the accuracy and efficiency of recommendations.

[0086] Finally, by comparing the similarity between the object nodes in the most relevant community and the user portrait to be recommended, the object node with the highest similarity is selected as the closest enterprise, and its corresponding enterprise problem solution is used as the final recommendation result. This method not only takes into account the current problems faced by users, but also fully combines the user's portrait characteristics to achieve personalized recommendations, which helps to improve user satisfaction and usage experience.

[0087] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, a user solution recommendation method based on enterprise user portrait analysis as described in any of the foregoing items is implemented.

[0088] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the method for recommending user solutions based on enterprise user portrait analysis as described in any of the foregoing items is implemented.

[0089] The above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. The protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A user solution recommendation method based on enterprise user portrait analysis, characterized in that: The recommended method comprises the following steps: S1. Determine the target user to be recommended, collect the first enterprise information of the target user to be recommended and the second enterprise information of the historical user who has formed a solution, and collect the enterprise problem solutions corresponding to the historical user; S2. Forming a to-be-recommended user portrait and a historical user portrait based on the first enterprise information and the second enterprise information; S3, vectorizing multiple historical user portraits as object nodes, vectorizing multiple enterprise problem solutions as result nodes, establishing a first edge based on the object nodes and the result nodes, and constructing a second edge based on multiple result nodes, wherein the set of the first edge, the second edge, the object node, and the result node constitute a hypergraph network; S4, performing community division on the hypergraph network; S5. Obtain the question input by the user and convert it into a user question vector, and search the community most relevant to the user question vector from the hypergraph network; S6. Compare the similarity between the object node in the community most related to the user problem vector and the user portrait to be recommended, select the object node with high similarity as the closest enterprise, and its corresponding enterprise problem solution as the solution recommendation result.

2. The user solution recommendation method based on enterprise user portrait analysis according to claim 1 is characterized in that: Information data of target users to be recommended and historical users for whom solutions have been formed are collected through database export, web crawler selection, and manual input.

3. The user solution recommendation method based on enterprise user portrait analysis according to claim 2 is characterized in that: The step of forming a to-be-recommended user profile and a historical user profile based on the first enterprise information and the second enterprise information specifically includes: Constructing an enterprise portrait model, which consists of five dimensions: basic attributes of the enterprise, business scope of the enterprise, main products of the enterprise, internal and external connections of the enterprise, and industry evaluation of the enterprise; Extract keywords from the first enterprise information as feature items, and fill the feature items into five corresponding dimensions of the enterprise's basic attributes, enterprise business scope, enterprise's main products, enterprise's internal and external connections, and enterprise's industry evaluation in the enterprise portrait model to form a user portrait to be recommended; The keywords in the second enterprise information are extracted as feature items, and the feature items are filled into the five corresponding dimensions of the basic attributes of the enterprise, the business scope of the enterprise, the main products of the enterprise, the internal and external connections of the enterprise, and the industry evaluation of the enterprise in the new enterprise portrait model to form a historical user portrait.

4. The user solution recommendation method based on enterprise user portrait analysis according to claim 2 is characterized in that: Step S3 specifically includes: Extract key sentences from the enterprise problem solution, and vectorize the key sentences and the enterprise problem solution as result nodes; Based on the correspondence between historical users and enterprise problem solutions, a first edge is established between an object node and a result node where the correspondence exists; Calculate the similarity between two key sentences in all pairwise result nodes. If the similarity between any two key sentences is greater than the first threshold, establish a second edge between the two result nodes, and the value of the similarity is the information of the second edge. A hypergraph network is constructed according to all the first edges and the second edges, and the hypergraph network includes result nodes and object nodes.

5. The user solution recommendation method based on enterprise user portrait analysis according to claim 4 is characterized in that: The Louvain algorithm is used to divide the hypergraph network into communities, and the divided communities contain a certain number of first edges, second edges, result nodes, and object nodes.

6. The user solution recommendation method based on enterprise user portrait analysis according to claim 5 is characterized in that: The question input by the user is obtained and converted into a user question vector, and the community most relevant to the user question vector is searched from the hypergraph network, specifically including: Use vector tools to vectorize the questions input by users, and convert the questions input by users into user question vectors; The similarity between the user question vector and all result nodes in each community in the hypergraph network is compared and calculated in turn to find the community most relevant to the user question vector.

7. The user solution recommendation method based on enterprise user portrait analysis according to claim 6 is characterized in that: In any community, if the number of result nodes that meet the similarity requirement with the user question vector is greater than the second threshold, it means that the community is the community most relevant to the user question vector.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the user solution recommendation method based on enterprise user portrait analysis as described in any one of claims 1 to 7 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for recommending user solutions based on enterprise user portrait analysis as described in any one of claims 1 to 7 is implemented.

Images