Cause Analysis of Hot Work Accidents Based on Text Mining and Deep Learning

This research combines deep learning and text mining to automatically analyze causes of hot work accidents in China's petrochemical industry, where such accidents account for over 50% of major chemical incidents. The method uses BERT-based classification to predict accident cause categories and TF-IDF with keyword importance scoring to identify key causal factors including lack of gas detection, uncleaned combustibles, and regulatory violations. The study directly supports PSM Incident Investigation by automating the extraction and categorization of root causes from large volumes of unstructured accident reports, providing safety managers with data-driven guidance for formulating targeted prevention strategies specific to hot work operations.