Using data science to predict firemen interventions: a case study

In the recent years, fire departments started to build databases containing detailed information about their interventions during fires, road accidents, and other types of incidents. Their goal is to invest this information using data analysis methods in order to better understand the trends of certain events. This could help them enhance the management of their allocated resources, which leads to a reduction in the operational costs, increase in efficiency and the overall intervention speed. Therefore, in this research paper, we investigate the possibility of predicting future incidents using machine learning algorithms that are trained on a set of data containing information on almost 200,000 interventions that happened during the last 6 years. These data, provided by the fire department in the region of Doubs, France, were not sufficient to detect patterns. Thus, we have imported additional information from external resources that we thought it would improve the accuracy of the predictions. Finally, we tested multiple machine learning algorithms and we compared their results, aiming to determine which algorithm performs better. The results look promising as we were able to predict the number of interventions for each 3 hours block for a whole year, with an acceptable error margin.