Research On Agricultural Online Information Resource Recommendation System Based On Knowledge Graph

With the gradual penetration of big data technology into agricultural production and the continuous expansion of agricultural knowledge databases,farmers can search for crop cultivation information and sales information through online searches.When searching for agricultural information through traditional keyword search and similar words,the online search results will be mixed with a lot of redundant,inaccurate and irrelevant data information.As a result,farmers need to spend a lot of time and energy to find useful answers in search results,which greatly reduces the user experience.Therefore,this article starts with the construction of knowledge graphs of agricultural-related databases,and focuses on the related technologies of the agricultural online resource recommendation system.The main research work of the thesis is as follows:(1)Construction of agricultural knowledge map.According to the organizational structure of the knowledge graph,the knowledge graph construction of the agricultural online resource recommendation system is divided into two parts: ontology layer construction and entity layer construction.The construction process needs to complete data collection,data cleaning,ontology layer construction,entity layer construction and knowledge graph formation.At the same time,using python data processing technology combined with the generated data wrapper to capture soil type data,crop diseases and insect pest data and agricultural trade data respectively,and use the native map data storage tool Neo4 j to store the graph database of agricultural knowledge graphs.(2)Propose a matrix factorization algorithm based on user preferences.By extracting the user’s preference for the entity information in the knowledge graph,and then controlling the communication intensity of the entity,the communication entity is always within the range of user preference.A comparative analysis of multi-attribute based classification,similarity measurement based on meta-graph normalization,and entity recommendation methods based on user preference dissemination are carried out.The experimental results show that the preference method has the best recommendation effect compared to the other two methods,with the smallest The F1 value is 87.54%.The matrix factorization algorithm is used to realize the purpose of connecting the resource database and the question sentences raised by users.The impact of item-based,user-based and matrix-based decomposition methods on the recognition accuracy of the recommendation system is compared and analyzed.Experiments show that matrix decomposition recommendations based on user preferences The algorithm has the best F1 value,and the F1 value reaches 0.896,0.827 and 0.703 in the three types of data sets.(3)Designed and implemented a recommendation system for agricultural online information resources based on knowledge graphs.The system is divided into an interface display module,a "user-answer" processing module,and a data building module,which can support crop planting and sales information query,and accurately recommend answer information in the agricultural field for farmers.The K-means clustering method,multi-feature fusion method and the method used in this article are compared and analyzed in the system recommendation accuracy and time delay respectively.The experimental results show that the method in this paper has the best recommendation effect,and the recommendation accuracy rate can reach 96%,the time delay is only 2s.