Improving the Location of Roadside Assistance Resources Through Incident Forecasting.

This paper presents a solution for a real world roadside assistance problem. Roadside assistance companies must allocate their specialised resources to minimize the operating cost associated with servicing when incidents occur. In this process, the location of these resources plays an important role. Therefore, this work proposes a study on the forecasting of incidents and their impact on the location of resources and operating costs. To do this, we have built a machine learning model competition enriched with new features drawn from traditional time series methods and external data such as weather, holidays, and client portfolios. The results show a significant reduction in operating costs thanks to the forecasting of incidents.