로지스틱 회귀분석을 활용한 건축물 해체공사 사고 예측모형 개발

초록

Amid the surging demolition demand due to aging buildings, a quantitative prevention system is urgently needed to lower the high fatality rate of 21.4%, approximately five times the general construction average. This study developed an accident severity prediction model using binary logistic regression on 290 demolition accidents reported to the Construction Safety Management Integrated Information System (CSI) from 2019 to 2025. The analysis showed that temperature and project scale decisively impact lethality. During heatwaves (≥33 °C), the fall fatality probabilities surged 32-fold, and 'crushed' accidents proved the most fatal with an odds ratio of 85.09. Specifically, a scenario of 'private projects, under 20 workers, and heatwaves' yielded a 97.27% predicted crushing fatality rate. The model showed a high classification accuracy of 84.1% for fatal accidents. Consequently, this study recommends climate-adaptive safety regulations, including mandatory work suspensions during heatwaves, and targeted inspections for high-risk scenarios. This research will significantly enhance site safety by presenting a probabilistic model that reflects the demolition-specific risk factors.

키워드