Knowledge Graph-based Knowledge Recommendation Technical Research And Implementation For Apple Diseases And Insect Pests

Currently,the apple industry is an essential pillar of China’s fruit industry.China has the most prominent apple-growing area in the world,where apple diseases and insect pests have been an important factor affecting the apple’s yield and quality.When encountering pests and diseases,fruit farmers often spend a lot of time searching for relevant knowledge,delaying the best time to treat pests and diseases,and affecting apples’ the quality.Therefore,it has become a significant research problem to be solved urgently in apple agriculture to recommend to fruit farmers the knowledge of apple diseases and insect pests they need in time.When solving the problem of apple disease and insect pest recommendation,the traditional-based recommendation algorithms are often poor.This is because such algorithms fail to focus on the high-order relationships between users and pests.Meanwhile,the data interpretability of such algorithms is flawed.To address the problems above,this study is based on the Knowledge Graph technology,constructing a User Apple Disease and Insect Pest Knowledge Graph.Considering the characteristics of knowledge graph constructed,a Knowledge Awareness Attention(KAA)recommendation model is proposed to extract the potential relationships between user entities and pest entities in the graph.Finally,a recommendation system for apple diseases and insect pests is designed and implemented based on the proposed knowledge graph and recommendation model.The paper’s main work is summarized as follows:(1)Build the User Apple Disease and Insect Pest Knowledge Graph.The data sources of the proposed Knowledge Graph are professional books and related websites.The initially acquired data needs to remove redundant or duplicated data.The cleaned data is then processed using annotation tools to mark the entities,relationships,and attributes present in the data.This work has identified 3,500 user entities,57 insect pest and disease entities,and 163 pesticide entities,etc.Finally,the tagged data is stored in the Neo4 j database to form the User Apple Disease and Insect Pest Knowledge Graph.(2)Design and research of the recommendation model.Considering the characteristics of the User Apple Disease and Insect Pest Knowledge Graph,the KAA model based on the knowledge graph is designed.KAA mainly includes a encoding layer,a Trans R-based embedding layer,an attention fusion layer,and a prediction layer.The Trans R-based embedding layer realizes the low-rank representation of graph data,while the attention fusion layer connects different positional features and extracts information hidden in relational information in low-rank representations.The ablation experiments prove that the Trans R-based embedding layer and attention fusion layer play a critical role in KAA.Compared with ten popular recommendation algorithms such as K-NCR and Graph SAGE,KAA performs best in Precision@50,Recall@50,and NDCG@50.Compared with Graph SAGE,which ranked second,KAA improves by about 0.95%,4.00%,and 2.13% in Precision@50,Recall@50,and NDCG@50,respectively.Compared with baseline algorithms such as Trans D,Trans R,Trans H,and Trans E,the improvement of KAA in Recall@50 and NDCG@50 is more than7.00%.Meanwhile,compared with the number of parameters and FLOPs of other models,the number of model parameters and FLOPs of KAA are in the middle position.This shows that KAA has a certain balance of performance and efficiency.(3)Construct the recommendation system.Based on the User Apple Disease and Insect Pest Knowledge Graph and KAA,a recommendation system for apple diseases and insect pests is designed and implemented.The system clarifies the system permissions of different roles to meet the needs of different users,so the system is divided into a management system and a user system.The management system is mainly for the management of system resources,including user information management,apple information management,insect pest management,pesticide management and log management.This system has the highest authority for system management.The user system is mainly to manage the user’s information and provide users with personalized services,such as viewing the user’s historical history and visualizing the knowledge graph of apple diseases and insect pests.Finally,the system test verifies the accuracy,security,and stability of the system to provide various users with their individual requirements.