Research on Information Attribute Extraction and Association Rule Mining of Hazardous Chemical Accident Investigation Report

Abstract

Although there are many studies on the investigation reports of hazardous chemical accidents, the current discussion has usually focused on a module of the report according to the different research purposes and there are problems such as ignoring other textual information and lacking a complete interpretation of the report. Its textual description also makes the report low in value and coarse-grained exploration. Moreover, hazardous chemical accidents are a process that constantly evolves and develops over time, and complexity and diversity of their occurrence scenarios make it difficult to explore more details of the evolution process of accidents by single-stage “cause-effect” rule mining model. In view of the above limitations, this article has proposed a new idea to solve the problem: analyze the hazardous chemical accident investigation reports systematically from the perspective of the scenario, and obtain the entire process and staged scenario representation theoretical frameworks applicable to massive accident investigation reports by scenario segmentation with human intervention as nodes and element extraction for report adequacy, and then according to the research purpose and demands, some elements in the framework can be selected to excavate meaningful and appropriate rules of whole-process and staged scenarios of hazardous chemical accidents by mining model combining rough sets and association rules. The suggestions provided in the study provide a way to explore the causes and knowledge behind l accidents from the macro and micro levels.