Abstract: In this article, we propose a method for mining disaster-causing factors in community fires based on information from the Internet. It collates dynamic data reflecting the lives of community residents from open maps, housing rental and sales information, merchant information, and relevant netizen comments, comprehensively use text mining, time series signal analysis, and other methods to study and judge the future fire situation and key data characteristics of the community. It also uses netizen comments to mine hidden community fire factors, and ultimately achieves the goal of dynamically identifying disaster-causing factors in community fires. The method can be used to assist the dynamic updating of community fire prevention strategies for fire management, departments and provide a new way of thinking for dynamic prevention and control of community fires.

Key words: community fire prevention, disaster-causing factors, multi-dynamic factors, risk assessment