Intelligent diagnostic system for rice diseases and pests based on knowledge graph

Objective  To conduct structured storage of complex and heterogeneous data information in the field of rice diseases and pests using knowledge graphs, establish semantic relationships between diseases and pests, and provide a theoretical basis for rice diseases and pests association retrieval and intelligent diagnosis.

Method  Firstly, a method of constructing a knowledge graph for rice diseases and pests was proposed. At the same time, a series of graph-based retrieval algorithms for rice diseases and pests were proposed for information mining, through introducing solar terms entities to achieve early warning of rice diseases and pests. Secondly, a knowledge reasoning method based on the combination of certainty factor (CF) model and knowledge graph was proposed to realize the intelligent diagnosis of rice diseases and pests by combining CF with the symptom of diseased plant.

Result  The accuracy rates of named entity recognition model were 0.92, 0.90, and 0.87 in disease and pest name and hazard symptom entities. Further, a knowledge graph of rice disease and pest domain including 1 972 entities and 5 226 entity relationships was constructed. Through the self-developed intelligent diagnosis system, case analysis was conducted and the test showed that the correct rate of the diagnosis algorithm reached 86.25%.

Conclusion  This study effectively solves the complexity and uncertainty of knowledge in data retrieval, early warning and diagnosis in the field of rice diseases and pests, and has a strong practical value and extension prospects.