A Smartphone-Based Hybrid Model for Real-Time Monitoring of Aggressive Driving Behavior using Multidimensional Dynamic Time Warping

Road safety is a significant global concern, with aggressive driving behavior being a leading cause of traffic accidents and fatalities worldwide. Accurate driver behavior prediction can enhance safety by enabling timely interventions and promoting responsible driving practices. This study proposes a hybrid model that uses multi-dimensional Dynamic Time Warping (DTW) and machine learning techniques to predict driver behavior. This model addresses the limitations of existing approaches by integrating DTW-derived features with traditional statistical features, hence, capturing both temporal alignment patterns and distributional driving data characteristics. DTW is employed to identify alignment patterns and quantify similarities across sequences of driving data. These DTW-derived features are augmented with the mean and variance of the corresponding sequences. This hybrid feature set is then used to train a random forest model. The performance of the proposed methodology is evaluated on a real-world dataset with accelerometer and gyroscope data. The hybrid model achieves an accuracy of up to 96%. This study also proposes an alert system with a feedback-based mechanism that uses the hybrid model to monitor and regulate driver behavior in both autonomous and manually driven vehicles.