Driver Behavior Assessment Based on Vehicle Data and Passenger Feedback Using Machine Learning

Evaluating driver behavior in fleet operations is essential for increasing driving safety, improving fuel economy, and optimizing overall driving performance. This study presents the development of a machine learning-based subjective evaluation model designed to assess driver behavior during acceleration and deceleration maneuvers, considering different speed ranges, pedal inputs, and perceived levels of aggressiveness. A structured experimental design was applied, involving the creation of specific driving maneuvers and the formulation of targeted questions to capture subjective feedback on driver behavior. The model utilizes real-time data collected throughout driving maneuvers and a questionnaire designed to measure passengers' perceived levels of safety and comfort. The developed model aims to establish a reliable correlation between measurable driving parameters and perceived driving quality, allowing subjective experiences to be evaluated within an objective framework. This approach has the potential to contribute to personalized driving experiences. Future studies aim to adapt this model to different driving conditions and vehicle types, thereby expanding its applicability.