Scientific research hotspot recommendation method based on knowledge graph

Abstract

The invention provides a scientific research hotspot recommendation method based on a knowledge graph, and the method comprises the steps: firstly, carrying out the structural representation of the information in the scientific research field through the construction of the knowledge graph, and improving the query performance and relevance of the information; and secondly, by adopting a knowledge graph embedding technology, the entities and the relationships are embedded into a low-dimensional vector space, so that efficient vector calculation is realized, and the efficiency of a recommendation algorithm is improved. In addition, through personalized recommendation and mixed recommendation algorithms, the interest and behavior data of the user are combined, and the recommendation accuracy and diversity are improved. And finally, through evaluation and optimization of the recommendation result, the performance of the recommendation system is continuously iteratively improved, and the user experience is improved.

Description

Technical Field

[0001] The present invention specifically relates to a method for recommending scientific research hotspots based on knowledge graphs. Background Art

[0002] Recommendation systems have been widely used in many scenarios in real life, especially personalized recommendation systems, which have seen increasing research and implementation. However, traditional recommendation systems still face some challenges, such as data sparsity, cold start problems, and insufficient interpretability of recommendation results. In order to solve these problems, researchers began to explore integrating knowledge graphs as auxiliary information into recommendation systems. This approach is called a knowledge graph-based recommendation system (KG-based Recommendation System, KGRS).

[0003] As a technology that connects information through semantic networks and reveals complex relationships between knowledge points, knowledge graphs have been widely used in many disciplines in recent years. By constructing structured representations of entities and relationships, it can more accurately capture the relationship between users and items, as well as user preferences. Therefore, applying knowledge graphs to recommendation systems can not only improve the accuracy of recommendations, but also provide explainability for recommendation results.

[0004] In summary, a scientific research hotspot recommendation method based on knowledge graph is proposed. Summary of the invention

[0005] The purpose of the present invention is to provide a scientific research hotspot recommendation method based on knowledge graph in view of the deficiencies of the prior art. The scientific research hotspot recommendation method based on knowledge graph can well solve the above problems.

[0006] In order to achieve the above requirements, the technical solution adopted by the present invention is: to provide a method for recommending scientific research hotspots based on knowledge graph, and the method for recommending scientific research hotspots based on knowledge graph includes the following steps:

[0007] S1: Data collection and preprocessing steps; collect a large amount of scientific research data, including papers, patents, projects, and scholar information. In the data preprocessing stage, it is necessary to clean the data, remove duplicates and invalid information, ensure the accuracy and completeness of the data, and convert the data into a structured format suitable for knowledge graph construction;

[0008] S2: Perform the steps of constructing a knowledge graph; based on the preprocessed data, construct a knowledge graph in the field of scientific research. The construction of the knowledge graph includes entity recognition, relationship extraction and graph storage. The entities are papers, scholars, institutions, and research fields, and the relationships are paper citations, author affiliations, and project cooperation. A graph database is used to store the graph for efficient query and update.

[0009] S3: Entity and relationship annotation steps: annotate entities and relationships in the knowledge graph to improve the efficiency of subsequent algorithms. Annotations include entity type annotation and relationship type annotation, which helps the algorithm better understand the structure and semantics of the graph in subsequent processing.

[0010] S4: Steps for embedding knowledge graphs; embed entities and relations in the knowledge graph into low-dimensional vector space for efficient vector calculations. Assuming that entities e_h and e_t are connected by a relation r, the goal of the embedding model is to minimize the loss function L(e_h, r, e_t), where L is usually defined as square error or cross entropy. By optimizing the embedding model, a low-dimensional vector representation of each entity and relation is obtained.

[0011]

[0012] S5: Steps for identifying research hotspots: Based on the entities and relationships in the knowledge graph, identify current research hotspots, evaluate their popularity by calculating the attention paid to different research fields or entities, and use the associations in the graph to discover potential research trends and intersections;

[0013] S6: Steps for constructing user portraits: Based on the user's scientific research interests and behavior data, a user portrait is constructed. The user portrait includes the user's research field, scholars, papers, and projects that the user is interested in. Collaborative filtering or content-based recommendation methods are used to extract entities and relationships related to the user's interests from the knowledge graph.

[0014] S7: the step of generating a personalized recommendation candidate set. Based on the user profile and scientific research hotspots, a personalized recommendation candidate set is generated. The candidate set includes papers, scholars, and projects related to the user's interests. By calculating the similarity between the candidate set and the user profile, the recommended items that best meet the user's needs are screened out.

[0015] S8: Steps of performing a path-based recommendation algorithm; using the path information in the knowledge graph to make recommendations, users follow the papers, and based on the citation relationship of the papers and the author cooperation relationship path, recommend related papers, scholars or projects. Effective strategies are designed for path selection and extraction to ensure the accuracy and diversity of recommendations;

[0016] S9: Steps for hybrid recommendation algorithm fusion; combining multiple recommendation methods to improve recommendation results. Hybrid recommendation algorithms can complement the advantages and disadvantages of different methods, increase the accuracy and diversity of recommendations, and use weighted average, voting mechanism or deep learning model to perform fusion;

[0017] S10: Steps for evaluating and optimizing recommendation results; evaluate the recommendation results, including recall rate, accuracy, and user satisfaction indicators. According to the evaluation results, optimize the recommendation algorithm. The optimization includes adjusting algorithm parameters, improving knowledge graph construction methods, and enriching user portraits. Through iterative optimization, the performance of the recommendation system and user experience are continuously improved.

[0018] The advantages of this scientific research hotspot recommendation method based on knowledge graph are as follows:

[0019] (1) Data sparsity: Traditional recommendation systems usually rely on the user-item interaction matrix for recommendation, but this matrix is ​​often sparse because users usually interact with only a few items. By introducing the knowledge graph, items and their attribute information can be mapped into the graph to understand the relationship between items. In this way, even if some user-item interaction data is missing, it can be supplemented by the relationship in the graph, alleviating the data sparsity problem.

[0020] (2) Cold start problem: Traditional recommendation systems often have difficulty making effective recommendations for new users or new items due to the lack of sufficient historical interaction data. Recommendation systems based on knowledge graphs can use the rich information in the graph to model and recommend new users or new items. For example, for new users, similar users can be found through their social network information or interest tags, and recommendations can be made based on the preferences of these similar users; for new items, recommendations can be made based on their attribute information or associations with other items.

[0021] (3) Explanability of recommendation results: The recommendation results of traditional recommendation systems often lack explainability, and users find it difficult to understand why a certain item is recommended. Recommendation systems based on knowledge graphs can explain recommendation results by displaying the relationship sequence in the graph. For example, users can be shown the association path between the items they liked before and the current recommended items, thereby increasing the credibility and acceptance of the recommendation. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The same reference numerals are used in these drawings to represent the same or similar parts. The exemplary embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0023] Figure 1 A schematic diagram of a method for recommending scientific research hotspots based on a knowledge graph according to an embodiment of the present application is schematically shown. DETAILED DESCRIPTION

[0024] In order to make the objectives, technical solutions and advantages of the present application more clear, the present application is further described in detail below in conjunction with the accompanying drawings and specific embodiments.

[0025] In the following description, references to "one embodiment", "an embodiment", "an example", "an example", etc. indicate that the embodiment or example described in this way may include specific features, structures, characteristics, properties, elements or limitations, but not every embodiment or example necessarily includes the specific features, structures, characteristics, properties, elements or limitations. In addition, repeated use of the phrase "according to one embodiment of the present application" may refer to the same embodiment, but does not necessarily refer to the same embodiment.

[0026] For the sake of simplicity, certain technical features well known to those skilled in the art are omitted in the following description.

[0027] According to an embodiment of the present application, a method for recommending scientific research hotspots based on knowledge graph is provided, such as Figure 1 As shown, the following steps are included:

[0028] Step 1: Data collection and preprocessing

[0029] First, a large amount of scientific research data needs to be collected, including papers, patents, projects, and scholar information. These data usually come from different databases and document management systems. In the data preprocessing stage, the data needs to be cleaned, duplicates and invalid information removed, and the accuracy and completeness of the data needs to be ensured. At the same time, the data is converted into a structured format suitable for knowledge graph construction, such as a triple (entity-relationship-entity) format.

[0030] Step 2: Build a knowledge graph

[0031] Based on the preprocessed data, a knowledge graph in the field of scientific research is constructed. The construction of the knowledge graph includes entity recognition, relationship extraction, and graph storage. Entities can be papers, scholars, institutions, research fields, etc., and relationships can be paper citations, author attribution, project cooperation, etc. Use a graph database (such as Neo4j) to store the graph for efficient query and update.

[0032] Step 3: Entity and Relationship Annotation

[0033] Label the entities and relationships in the knowledge graph to improve the efficiency of subsequent algorithms. Labeling includes entity type labeling (such as papers, scholars) and relationship type labeling (such as citations, collaborations). This step helps the algorithm better understand the structure and semantics of the graph in subsequent processing.

[0034] Step 4: Knowledge Graph Embedding

[0035] Embed entities and relations in the knowledge graph into a low-dimensional vector space for efficient vector calculations. Use models such as TransE, DistMult, or ComplEx for embedding. Assuming that entities e_h and e_t are connected by a relationship r, the goal of the embedding model is to minimize the loss function L(e_h,r,e_t), where L is usually defined as square error or cross entropy. By optimizing the embedding model, a low-dimensional vector representation of each entity and relationship is obtained.

[0036]

[0037] Step 5: Identification of research hotspots

[0038] Based on the entities and relationships in the knowledge graph, we can identify current research hotspots. We can evaluate the popularity of different research fields or entities by calculating their attention (such as the number of paper citations, project funding support, etc.). At the same time, we can use the associations in the graph to discover potential research trends and intersections.

[0039] Step 6: User portrait construction

[0040] Build user profiles based on the user's scientific research interests and behavior data. User profiles include the user's research field, scholars, papers, projects, etc. Use collaborative filtering or content-based recommendation methods to extract entities and relationships related to user interests from the knowledge graph.

[0041] Step 7: Generate personalized recommendation candidate sets

[0042] Generate personalized recommendation candidate sets based on user profiles and research hotspots. The candidate sets include papers, scholars, projects, etc. related to user interests. By calculating the similarity between the candidate sets and user profiles, select the recommended items that best meet user needs.

[0043] Step 8: Path-based recommendation algorithm

[0044] Use the path information in the knowledge graph for recommendation. For example, if a user is interested in a certain paper, related papers, scholars or projects can be recommended based on the paper's citation relationship, author cooperation relationship and other paths. The selection and extraction of paths requires the design of effective strategies to ensure the accuracy and diversity of recommendations.

[0045] Step 9: Hybrid recommendation algorithm fusion

[0046] Combine multiple recommendation methods (such as collaborative filtering, content-based recommendation, path-based recommendation, etc.) to improve the recommendation effect. Hybrid recommendation algorithms can complement the advantages and disadvantages of different methods and increase the accuracy and diversity of recommendations. Use weighted average, voting mechanism or deep learning model for fusion.

[0047] Step 10: Evaluation and optimization of recommendation results

[0048] Evaluate the recommendation results, including recall rate, accuracy, user satisfaction and other indicators. According to the evaluation results, optimize the recommendation algorithm. Optimization includes adjusting algorithm parameters, improving knowledge graph construction methods, enriching user portraits, etc. Through iterative optimization, continuously improve the performance of the recommendation system and user experience.

[0049] According to one embodiment of the present application, the method for recommending scientific research hot spots based on the knowledge graph

[0050] According to an embodiment of the present application, the method for recommending scientific research hot spots based on the knowledge graph specifically includes the following steps:

[0051] 1. Data Collection and Preprocessing

[0052] To implement the operation:

[0053] Data collection: Collect data such as papers, patents, projects, and scholar information from major scientific research databases, document management systems, academic forums, etc. to ensure the comprehensiveness and diversity of the data.

[0054] Data cleaning: remove duplicates, invalid information, erroneous data, etc. to ensure the accuracy and completeness of the data.

[0055] Structural processing: Convert the data into a structured format suitable for knowledge graph construction, such as triples (entity-relationship-entity). This step may require the use of natural language processing (NLP) technology for entity recognition and relationship extraction.

[0056] 2. Building a Knowledge Graph

[0057] To implement the operation:

[0058] Entity recognition: Identify entities in the collected data, such as papers, scholars, institutions, etc. This step can be performed using rule-based methods, machine learning methods, or deep learning methods.

[0059] Relationship extraction: Identify the relationships between entities, such as paper citation relationships, author collaboration relationships, etc. This step can also be performed using a variety of methods, including template-based methods, supervised learning methods, etc.

[0060] Graph storage: Use a graph database (such as Neo4j) to store graph data for efficient query and update. When storing, you need to consider the scale and performance requirements of the graph and choose an appropriate storage strategy and indexing mechanism.

[0061] 3. Entity and Relationship Annotation

[0062] To implement the operation:

[0063] Annotation strategy design: Design appropriate annotation strategies based on the characteristics and needs of the scientific research field, including entity type annotation (such as papers, scholars, etc.) and relationship type annotation (such as citations, collaborations, etc.).

[0064] Selection of annotation tools: Choose the appropriate annotation tool for annotation work. You can use existing annotation tools for secondary development, or develop your own annotation tools according to your needs.

[0065] Annotation quality control: Perform quality control on the annotation results to ensure the accuracy and consistency of the annotations. Quality control can be performed using methods such as manual review and cross-validation.

[0066] 4. Knowledge Graph Embedding

[0067] To implement the operation:

[0068] Embedding model selection: According to the characteristics and requirements of the knowledge graph, select a suitable embedding model for embedding. Commonly used embedding models include TransE, DistMult, ComplEx, etc.

[0069] Model training and optimization: Use the collected graph data to train the model and optimize the model through optimization algorithms (such as gradient descent). During the training process, the model parameters need to be continuously adjusted to improve the embedding effect.

[0070] Embedding result evaluation: Evaluate the embedding results, including performance evaluation on tasks such as link prediction and triple classification. Based on the evaluation results, further adjust and optimize the embedding model.

[0071] 5. Identification of scientific research hotspots

[0072] To implement the operation:

[0073] Attention calculation: Based on the entities and relationships in the knowledge graph, the attention of different research fields or entities is calculated. Attention can be measured by indicators such as the number of paper citations and project funding support.

[0074] Hotspot identification algorithm: Design a suitable hotspot identification algorithm to identify current scientific research hotspots from entities and relationships with high attention. The algorithm can be designed based on statistical methods, machine learning methods or deep learning methods.

[0075] Hotspot verification and update: Verify and update the identified scientific research hotspots. Verification can be carried out through manual review, expert evaluation and other methods to ensure the accuracy and timeliness of the hotspots. At the same time, the hotspot library is continuously updated and expanded based on new data and research results.

[0076] 6. User portrait construction

[0077] To implement the operation:

[0078] User interest mining: Mining users’ interests based on their scientific research interests and behavior data. Users’ interests can be mined by analyzing their reading records, search records, social relationships and other data.

[0079] User portrait generation: Generate user portraits based on mined user interests. User portraits include the user's research field, scholars, papers, projects, etc. Portrait generation can be performed using collaborative filtering, content-based recommendation, and other methods.

[0080] Profile update and maintenance: As user interests and behaviors change, the user profile is continuously updated and maintained. The profile can be updated and adjusted by regularly collecting and analyzing user data.

[0081] 7. Personalized Recommendation Candidate Set Generation

[0082] To implement the operation:

[0083] Candidate set generation strategy: Design a suitable candidate set generation strategy based on user portraits and scientific research hotspots. The strategy can be designed based on multiple dimensions such as user interests, hot spot attention, entity relationships, etc.

[0084] Candidate set screening and optimization: Extract entities and relationships related to user interests from the knowledge graph to generate candidate sets. Use filtering algorithms, sorting algorithms, etc. to screen and optimize the candidate sets to ensure the accuracy and diversity of the recommendation results.

[0085] Candidate set evaluation and feedback: Evaluate the generated candidate set, including indicators such as recall rate and accuracy. At the same time, collect user feedback on the recommendation results to further adjust and optimize the recommendation algorithm.

[0086] Step 8 specifically includes:

[0087] Path similarity calculation: Given two entities e1 and e2, and a path p between them, consisting of a series of relations r1, r2, ..., rn, the similarity of path p is defined as:

[0088]

[0089] Among them, sim(ri) represents the similarity of relation ri, which is calculated by the embedding vector obtained through training;

[0090] Entity similarity calculation: For two entities e1 and e2, find all possible paths between them and calculate the sum of the similarities of these paths as the similarity between them:

[0091]

[0092] Among them, paths(e1,e2) represents the set of all possible paths between e1 and e2, and w(p) represents the weight of path p, which is set according to factors such as path length and relationship type.

[0093] Recommendation strategy: Based on the calculated entity similarity, select several entities that are most similar to the entity e that the user is currently paying attention to as the recommendation results.

[0094] Step nine specifically includes:

[0095] Weighted average formula:

[0096] Assume that there are n different recommendation algorithms, and the recommendation score of each algorithm for user u to item i is rec_k(u,i) and k=1,2,...,n. Assign a weight β_k to each algorithm, satisfying ∑β_k=1, and use the following formula to calculate the final recommendation score:

[0097]

[0098] Weight distribution: The weight distribution is determined based on the performance of the algorithm, user preferences, and business scenario factors. In practical applications, weight distribution is optimized through experiments or machine learning methods;

[0099] Recommendation list generation: Based on the calculated final recommendation score, select the items with the highest scores as the recommendation list and present it to the user.

[0100] 10. Evaluation and Optimization of Recommendation Results

[0101] To implement the operation:

[0102] Selection of evaluation indicators: According to the needs and characteristics of the scientific research hotspot recommendation, select appropriate evaluation indicators for evaluation. Commonly used evaluation indicators include recall rate, accuracy, user satisfaction, diversity, etc.

[0103] Evaluation method design: Design a suitable evaluation method for evaluation. You can use offline evaluation, online evaluation, A / B testing and other methods to evaluate. The accuracy and consistency of the data must be ensured during the evaluation process.

[0104] Optimization strategy formulation and implementation: Based on the evaluation results, formulate appropriate optimization strategies for implementation. Optimization strategies may include adjusting algorithm parameters, improving knowledge graph construction methods, enriching user portraits, etc. Through iterative optimization, the performance and user experience of the recommendation system are continuously improved.

[0105] The above-mentioned embodiments only represent several implementation methods of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the present invention. It should be pointed out that, for a person skilled in the art, several modifications and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the claims.

Claims

1. A method for recommending scientific research hotspots based on knowledge graph, characterized in that: Including the following steps: S1: Data collection and preprocessing steps; Collect a large amount of scientific research data, including papers, patents, projects, and scholar information. In the data preprocessing stage, it is necessary to clean the data, remove duplicates and invalid information, ensure the accuracy and completeness of the data, and convert the data into a structured format suitable for knowledge graph construction; S2: Perform the steps of building a knowledge graph; Based on the preprocessed data, a knowledge graph in the field of scientific research is constructed. The construction of the knowledge graph includes entity recognition, relationship extraction and graph storage. The entities are papers, scholars, institutions, and research fields, and the relationships are paper citations, author affiliations, and project collaborations. A graph database is used to store the graph for efficient query and update. S3: Steps for entity and relationship annotation; Label the entities and relationships in the knowledge graph to improve the efficiency of subsequent algorithms. The labeling includes entity type labeling and relationship type labeling, which helps the algorithm better understand the structure and semantics of the graph in subsequent processing; S4: Steps for knowledge graph embedding; Embed the entities and relations in the knowledge graph into a low-dimensional vector space for efficient vector calculation. Assuming that entities e_h and e_t are connected by a relation r, the goal of the embedding model is to minimize the loss function L(e_h,r,e_t), where L is usually defined as the square error or cross entropy. By optimizing the embedding model, a low-dimensional vector representation of each entity and relationship is obtained. S5: Steps for identifying research hotspots; Based on the entities and relationships in the knowledge graph, we can identify current research hotspots, evaluate the popularity of different research fields or entities by calculating their attention, and use the associations in the graph to discover potential research trends and intersections. S6: Steps for constructing user portraits; Build user profiles based on the user's scientific research interests and behavior data. The user profiles include the user's research field, scholars, papers, and projects that they follow. Use collaborative filtering or content-based recommendation methods to extract entities and relationships related to the user's interests from the knowledge graph. S7: a step for generating a personalized recommendation candidate set; Generate personalized recommendation candidate sets based on user profiles and research hotspots. The candidate sets include papers, scholars, and projects related to user interests. By calculating the similarity between the candidate sets and user profiles, select the recommended items that best meet user needs. S8: performing a path-based recommendation algorithm step; Use the path information in the knowledge graph to make recommendations. For papers that users are interested in, based on the citation relationship of the papers and the path of the author's cooperation relationship, recommend related papers, scholars or projects. Design effective strategies for path selection and extraction to ensure the accuracy and diversity of recommendations. S9: Steps for performing hybrid recommendation algorithm fusion; Combining multiple recommendation methods to improve the recommendation effect. Hybrid recommendation algorithms can complement the advantages and disadvantages of different methods, increase the accuracy and diversity of recommendations, and use weighted average, voting mechanism or deep learning model for fusion; S10: Steps for evaluating and optimizing the recommendation results; The recommendation results are evaluated, including recall rate, accuracy, and user satisfaction indicators. Based on the evaluation results, the recommendation algorithm is optimized. The optimization includes adjusting algorithm parameters, improving knowledge graph construction methods, and enriching user portraits. Through iterative optimization, the performance of the recommendation system and user experience are continuously improved.

2. The method for recommending scientific research hot spots based on knowledge graph according to claim 1, characterized in that: Step S1 specifically includes: Data collection: Collect data such as papers, patents, projects, and scholar information from scientific research databases, document management systems, and academic forums to ensure the comprehensiveness and diversity of the data; Data cleaning: remove duplicates, invalid information, and erroneous data to ensure data accuracy and completeness; Structured processing: Convert data into a structured format suitable for knowledge graph construction. The structured format is a triple, namely entity-relationship-entity form, and use natural language processing technology for entity recognition and relationship extraction.

3. The method for recommending scientific research hot spots based on knowledge graph according to claim 1, characterized in that: Step S2 specifically includes: Entity recognition: Identify entities in the collected data, including papers, scholars, and institutions. This step is performed using rule-based methods, machine learning methods, or deep learning methods. Relation extraction: Identify the relationships between entities, including paper citation relationships and author collaboration relationships. This step is performed using supervised learning methods. Graph storage: Use a graph database to store graph data for efficient query and update. When storing, select the appropriate storage strategy and indexing mechanism based on the scale and performance requirements of the graph.

4. The method for recommending scientific research hot spots based on knowledge graph according to claim 1, characterized in that: Step S3 specifically includes: Annotation strategy design: Design appropriate annotation strategies based on the characteristics and needs of the scientific research field, including entity type annotation and relationship type annotation; Annotation quality control: Perform quality control on the annotation results to ensure the accuracy and consistency of the annotations, and use cross-validation methods for quality control; Step S4 specifically includes: Embedding model selection: Select an embedding model for embedding based on the characteristics and requirements of the knowledge graph; Model training and optimization: Use the collected graph data to train the model and optimize the model through optimization algorithms. During the training process, continuously adjust the model parameters to improve the embedding effect. Embedding result evaluation: Evaluate the embedding results, including performance evaluation on link prediction and triple classification tasks. Based on the evaluation results, further adjust and optimize the embedding model.

5. The method for recommending scientific research hot spots based on knowledge graph according to claim 1, characterized in that: Step S5 specifically includes: Attention calculation: Based on the entities and relationships in the knowledge graph, the attention of different research fields or entities is calculated. The attention is measured by the number of paper citations and project funding support indicators; Hotspot identification algorithm: Identify current research hotspots from entities and relationships with high attention. The algorithm is based on statistical methods, machine learning methods or deep learning methods; Hotspot verification and update: Verify and update the identified scientific research hotspots to ensure their accuracy and timeliness, and continuously update and expand the hotspot database based on new data and research results; Step S6 specifically includes: User interest mining: Mining users’ interests based on their scientific research interests and behavior data, and mining their interests by analyzing their reading records, search records, and social relationship data; User portrait generation: Generate user portraits based on mined user interests. User portraits include the user's research field, scholars, papers, and projects that the user is interested in. The portraits are generated using collaborative filtering and content-based recommendation methods. Portrait update and maintenance: As user interests and behaviors change, user portraits are continuously updated and maintained, and portraits are updated and adjusted by regularly collecting and analyzing user data.

6. The method for recommending scientific research hot spots based on knowledge graph according to claim 1, characterized in that: Step S7 specifically includes: Candidate set generation strategy: Design appropriate candidate set generation strategies based on user portraits and scientific research hotspots. The strategies are designed based on multiple dimensions such as user interests, hot spot attention, and entity relationships. Candidate set screening and optimization: Extract entities and relationships related to user interests from the knowledge graph, generate candidate sets, and use filtering algorithms, sorting algorithms, etc. to screen and optimize the candidate sets to ensure the accuracy and diversity of recommendation results; Candidate set evaluation and feedback: Evaluate the generated candidate sets, including recall and accuracy indicators, and collect user feedback on the recommendation results in order to further adjust and optimize the recommendation algorithm.

7. The method for recommending scientific research hot spots based on knowledge graph according to claim 1, characterized in that: Step S8 specifically includes: Path similarity calculation: Given two entities e1 and e2, and a path p between them, consisting of a series of relations r1, r2, ..., rn, the similarity of path p is defined as: Among them, sim(ri) represents the similarity of relation ri, which is calculated by the embedding vector obtained through training; Entity similarity calculation: For two entities e1 and e2, find all possible paths between them and calculate the sum of the similarities of these paths as the similarity between them: Among them, paths(e1,e2) represents the set of all possible paths between e1 and e2, and w(p) represents the weight of path p, which is set according to factors such as path length and relationship type. Recommendation strategy: Based on the calculated entity similarity, select several entities that are most similar to the entity e that the user is currently paying attention to as the recommendation results.

8. The method for recommending scientific research hot spots based on knowledge graph according to claim 1, characterized in that: Step S9 specifically includes: Weighted average formula: Assume that there are n different recommendation algorithms, and the recommendation score of each algorithm for user u to item i is rec_k(u,i) and k=1,2,...,n. Assign a weight β_k to each algorithm, satisfying ∑β_k=1, and use the following formula to calculate the final recommendation score: Weight distribution: Weight distribution is determined based on algorithm performance, user preferences, and business scenario factors. In practical applications, weight distribution is optimized through experiments or machine learning methods; Recommendation list generation: Based on the calculated final recommendation score, select the items with the highest scores as the recommendation list and present it to the user.

Images