Predicting Fire Incidents in Industrial Facilities and Developing Customized Fire Insurance Products Using Data Mining

The recent increase in large-scale fires has led to significant loss of life and property, posing a serious threat to social safety. This study aims to analyze the factors and patterns of fire incidents in industrial facilities using data mining techniques to offer some insight to the development of customized fire insurance products. In particular, this study focuses on fire grades (fireproof structures and non-combustible materials), crucial factors in fire insurance design. Fire incidents at industrial facilities were analyzed using annual fire statistics (2019–2022) provided by the National Fire Agency of South Korea. Additionally, fire insurance subscription data from Korea Statistical Information Service was referenced to analyze current fire insurance enrollment status. For data analysis, predictive models such as Decision Tree (C5.0), CART (Gini), CHAID, and Random Forest were used, with Mean Absolute Error (MAE) and classification accuracy employed to verify prediction accuracy. While previous studies on fire prediction that primarily focused on time, place, and climate factors, this study set building structure and fire grade as primary factors in fire insurance design. This study contributes to the literature on fire risk prediction and fire insurance studies by providing an essential theoretical basis for fire prevention and risk management in industrial facilities.