Artificial Intelligence-Based Perceived Motorcycle Risk Prediction in Bangladesh’s Urban Driving Environment

Bangladesh has the highest fatality rates for motorcycle accidents worldwide, accounting for the largest portion of the total road crashes, and the rate has been increasing for the last couple of years. In 2022, data from the BUET Accident Research Institute (ARI) revealed that, because of accessibility, affordability, and ride-sharing usage, motorcycles account for 62% of the total number of vehicles plying the roads, with 26 accidents for every 10,000 motorcycles. So, it is imperative to investigate motorbike accident precursors, their contribution to endangerment, risk prediction and devise necessary policy implications. In this study, perceived risk data were collected through offline and online questionnaire surveys from 1559 respondents. Demographic data, like age, gender, residence division; usage criteria of motorbikes, occupation, ride-sharing app usage, and rating on the perceived risk of 38 precursors of motorbike accident in Bangladesh’s context were collected. Later precursors were clustered into 10 major combined attributes: biker’s driving behavior, motorbike condition, safety status, weather environment, pavement condition, driving environment, sign marking and lighting, traffic control, traffic movement, and pedestrian activity. Random forest algorithm has been used to predict perceived risk due to driving environment with 70% accuracy. Lastly, different contour maps for features’ correlation, feature importance, heat map, deployment of results in public server for user interface, and policy implications have been demonstrated in this study. In summary, the proposed prediction tools will have a significant prospect for accident analysis and prevention in the urban context of any developing nation.