Driving Risk prediction over a repetitive driving pattern

Road traffic accidents remain a major global issue, yet professional drivers face a distinct and often underestimated risk: long-term exposure to repetitive routes. To address this, we conducted a longitudinal study with 54 participants over a 12-day protocol, performing 4 consecutive laps during each day on a simulator on a repetitive track to replicate routine shifts. The resulting dataset fuses driving, eye-tracking metrics, and physiological features. An ML model was trained to predict the risk level of a lap with data from the previous one. The model was validated with a K-Fold approach, by achieving a mean accuracy of 91%.