Measurement of Drivers' Confidence Levels in Left-Turn Decisions Using Machine Learning

Decision confidence in driving, especially in complex operations such as left turns, is crucial for understanding human cognitive behavior and improving driving safety. At present, the lack of confidence quantification methods for dynamic real-time driving tasks limits the development of intelligent driving systems. This study proposes a method based on machine learning, which for the first time utilizes driving performance data (such as speed, acceleration, steering wheel angle) to measure driver confidence levels in left turn decisions. In a simulated driving environment, data from 33 participants were used to train seven machine learning models. Among them, the stacked classifier performed the best within the <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$0-5$</tex> second time window after stepping on the accelerator, with an accuracy of 90.97% and an <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$F 1$</tex> score of 90.91%, better than the <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{5 - 1 0}$</tex> second time window. The study identifies the standard deviation of forward direction position, maximum speed and mean speed as key features for confidence measurement. These findings provide the possibility for real-time assessment of driver confidence and have significant potential for improving human-machine interaction in intelligent driving systems and enhancing road safety.