Modeling speeding behaviour of light commercial vehicle drivers using binary logistic regression and machine learning techniques

Speeding behaviour is one of the most common hazardous driving practices contributing to road accidents globally. However, there is limited research on the determinants influencing speeding behaviour among Light Commercial Vehicle (LCV) drivers, particularly in developing nations such as India. This study aims to develop a predictive model and analyze the primary factors affecting speeding behaviour among LCV drivers in India. While statistical techniques such as binary logistic regression showed moderate predictive capability, machine learning techniques, namely Support Vector Machine (SVM) and Artificial Neural Network (ANN), achieved superior performance. Among the developed models, the linear kernel SVM yielded the best results with an accuracy of 80%, F1-score of 84%, and an AUC of 0.89. Variable importance analysis identified pressured delivery, marital status, education, driver age, smoking frequency, and vehicle age as key determinants. These findings can support policymakers in developing targeted strategies to reduce speeding among LCV drivers.