Abnormal Driving Behaviour Classification Using Machine Learning

Abstract

Smartphone sensor technology has paved the way for developing intelligent systems to detect and classify driving behaviour accurately. This paper presents a comprehensive study aimed at developing such a system by leveraging smartphone sensor data. The objective is to enhance road safety by accurately identifying abnormal driving behaviour. A Decision Tree (DT) based ML model has been proposed in this paper to detect and classify abnormal driving behaviours accurately into nine different classes. A unique combination of six parameters has been used to train and test the model, including vehicle speed read through the smartphone's GPS sensor, and pitch speed, roll speed, and acceleration on the X- and Y-axis read through the smartphone's inertial sensor. The results revealed that the DT-based ML model is highly accurate with precision, recall and F1-score of 0.944, 0.940 and 0.942, respectively, in classifying the abnormal driving behaviour. The developed system showcases ML algorithms' capabilities and exemplifies smartphone sensors' utility in creating intelligent transportation systems. By reducing the occurrence of accidents and fatalities, this affordable system holds tremendous potential to impact public safety significantly.