Are you a Good Driver?: A Data-driven Approach to Estimate Driving Style

In recent years, the study of aggressive driving behavior has reached interest due to its correlation with traffic accidents. Traffic accidents are considered the third cause of deaths in the United States. Prior research has reported the possibility to estimate aggressive driving style using in-vehicle data (e.g., acceleration, speeding, lane changes, among others). However, traffic violations have not yet been considered in the analysis of aggressive driving style. This paper proposes a model to estimate driver's aggressive driving style by considering aggressive events from in-vehicle data, and traffic violations data using a fuzzy logic model. In-vehicle data and GPS data from twenty-five drivers in different routes were collected, to generate a fuzzy logic model that captures aggressive events and traffic violations. We validate these results by comparing the results between the fuzzy logic model and human experts scores, showing an accuracy of 0.84 and a recall of 0.8974. Future work should consider to revise the rules and membership values to improve misclassification errors.