Prediction of road accidents in Cartagena, Colombia, using decision trees and association rules

Abstract

The main objective of this research is to predict the factors associated with the severity of road accidents in Cartagena (Colombia); the methodology is based on data mining techniques such as decision trees (J48) and association rules (support, confidence, Lift). The research was developed with 10,053 traffic accident records between 2016 and 2017, using the WEKA (Waikato Environment for Knowledge Analysis) software. In the analysis, the severity was defined as low risk (material damage), and high risk (injured and fatal victims), and its validation considered the 10-fold cross-sectional technique. Among the most significant results, it was evidenced that motorcyclists, cyclists are the most vulnerable road users, and motorcyclists between the ages of 20-39 are prone to road accidents with high severity. Finally, the road accident factors identified help promote countermeasures to improve the city’s road safety.