Spanish Road Fork Traffic Analysis and Modelling

This study focuses on the analysis of traffic data concerning road crashes, with the aim of extracting relevant knowledge in order to choose the location of intelligent road guardrails. To do so, historical data of the accidents in Spanish roads since 2011 have been gathered from the Direccion General de Trafico and analyzed afterwards. After a preliminary stage, where association rule mining algorithms were performed, this study focuses on Decision Tree classifiers to determine the relationships among the features in the accidents yearly dataset. These relationships are related to the intrinsic connections between features. Some interesting relationships have been found, specially for the T or Y shape road fork type. Future work will deploy a severity index using both the victims and vehicles, so relationships due to conditions and severity can be extracted. As long as the extracted rule set varies for each year, it might be useful as well to determine invariants in the trees, which is led as future work.