Local Outlier Factor와 의사결정나무 알고리즘을 활용한 이상사고 주요원인 분석

This study scientifically analyzes the main causes of abnormal accidents using datamining techniques based on text data contained in the accident report document. First, the textmining is used to build document-term matrix and the local outlier factor (LOF) is applied to discover abnormal accidents. Second, the main causes of abnormal accidents are hierarchically extracted through decision tree algorithm. As a case study, the chemical industry that handles various hazardous substances is conducted to retrieve abnormal accidents and to classify the main causes of abnormal conditions during the accident process. Also, we develop the checklist constructed by keywords included in abnormal accidents according to the process safety indicators. The proposed approach enables safety managers to apply text information contained in a lot of accident reports for safety management by extracting the main causes of abnormal accidents.