Anonymously forecasting the number and nature of firefighting operations
Jean-François Couchot, Christophe Guyeux, Guillaume Royer · 2019
Predicting the number and the type of operations by civil protection services is essential, both to optimize on-call firefighters in size and competence, to pre-position material and human resources... To accomplish this task, it is required to possess skills in artificial intelligence, which are not usually found in a medium-sized fire department. However, such a request may be mandated, for example from specialized companies or research laboratories. This mandate requires the transmission of potentially sensitive information relating to interventions which is not intended to be publicly available. The purpose of this article is to show that a machine learning tool can be deployed and provide accurate results, using a learning process based on anonymized data. Learning on real but anonymized data will be performed using extreme gradient boosting, and the performance of each anonymization will be compared on the number and of interventions per day, and their type.