Chimney fire prediction in IJsselland

In this paper, we develop a combined machine learning and statistical modelling process to predict chimney fire risk in IJsselland. We first use random forests and permutation importance techniques to find the explanatory variables. Secondly, we design a Poisson point process model and use logistic regression estimation to estimate the parameters. Afterwards, we verify the results using residuals. In all phases, we compare the results to the results from the Twente region, which is very similar to IJsselland. All the modelling is done on data collected by the Twente and IJsselland Fire Brigade.