Fire Risk Prediction Using Multi-Source Data: A case study in Humberside area

Abstract

Urban fires cause various problems to people and economy; thus, early detection and prevention of fires could save not only properties but also the people’s lives. Predicting fires is always a high priority for not only fire services but also everyone living in urban areas. In this study, we propose a model to integrate multiple data sources for identifying commercial buildings in the Humberside region that have a high risk of catching fire. This annual fire prediction model has the ability to fuse data from a local fire and rescue service along with some public datasets. We have compared prediction performance of several algorithms against AUC and average precision criteria and found that AdaBoost outperforms the others. The prediction results of AdaBoost are then used to identify risk levels of commercial properties. In comparison with the current fire risk assessment methods used by the fire and rescue service, this method significantly improves the accuracy.