Analysis of chemical production accidents in China: data mining, network modeling, and predictive trends
Yang Shi, Haitao Bian, Qingguo Wang, Yong Pan, Juncheng Jiang
Abstract
In recent years, China has experienced frequent chemical production accidents. This study collates 1900 reports of such incidents from 2012 to 2023, gathered from multiple sources. By employing association rule mining, a data mining technique, we analyzed the relationships between causative factors of these accidents and their patterns. This analysis revealed significant association rules characterized by high lift values, severe consequences, or previously under recognized patterns. Utilizing Gephi® software, we constructed a network model representing the causative factors of these accidents. Through centrality analysis of network nodes, we identified key factors contributing to these incidents. Additionally, we developed and validated a SARIMAX model using time series data of chemical production accidents, enabling predictions about future trends. The model's forecasts offer valuable insights for businesses in identifying periods with a higher likelihood of accidents. Conclusively, this comprehensive analysis and predictive modeling provide a critical framework for enhancing safety measures and proactive risk management in China's chemical production industry.