Every day, when firefighters respond to emergencies, they and the public face an unnecessary risk due to inadequate staffing. Having too many people stand-by costs a lot of money, on the other hand, having too few people stand-by leads to unnecessary safety risks. Therefore, for adequate staffing pur- poses, forecasting the number of incidents that each fire station has to handle is a very relevant question. In this paper, we develop models to create a good forecast for the number of incidents that each fire station in Amsterdam-Amstelland has to handle. Previous studies mainly focused on multiplicative models containing correction factors for the weekday and the time of the year. Our main contribution is to incorporate the influence of different weather conditions in the categories of wind, temperature, rain, and visibility. We show that an ensemble model has the best predictive performance. Rain and wind typically have a strong linear influence, while temperature mainly has a nonlinear influence.

, , ,
International Conference on Data Analytics

Legemaate, G., Bhulai, S., & van der Mei, R. (2019). Applied urban fire department incident forecasting. In Proceedings of IARIA Data Analytics (pp. 57–62).