Data mining model for food safety incidents based on structural analysis and semantic similarity

Food safety is of vital interest for public health and the stability of society. In this paper, we analyzed the characteristics of food safety incidents (FSIs), including spatial distribution, food categories, risk factors, and supply chain links, reported by mainstream media in China. Based on our analysis, we constructed a semantic template for text data related to FSIs. Furthermore, we introduced a multi-layer, multi-level semantic structure of rank (MMSS-Rank) algorithm to measure the similarity between collected food safety data and the semantic template. We then calculated the overall scores (i.e., text layer weight, semantic template weight, and keyword density matrix) and selected an appropriate threshold to determine the accuracy of the FSI data. Results showed that, compared with traditional methods, MMSS-Rank is an efficient and robust method for identifying large-scale FSI data with higher accuracy and recall rate.