Predicting Drivers’ Speeding Behaviour: A Case-based Reasoning Approach

To ensure successful road safety controls, several studies have attempted to propose models and theories that explain and predict drivers’ risky behaviour. However, existing research has several limitations including knowledge elicitation, bias, assumption, and interpretation issues. To help identify potential risky driving behaviours and avoid crashes, we have developed a case-based reasoning system that identifies speeding -one of the most common traffic violations. Our proposed system generates the expected value of a targeted traffic situation based on several similar previous cases. We consider four categories of factors, which include demographic information of the driver, vehicle information, road situation, and environmental factors. We examined four scenarios based on traffic crash data in China, and our case-based prediction system demonstrated its merits compared to the widely employed logistic regression model. Our findings will shed light on safety promotion strategies to decrease the rate of speeding and other traffic violations.