Motorised two-wheelers (TW) are integral part of India’s transportation systems and account for the highest proportion of road fatalities (37%). For improving the safety of TW riders, factors contributing to fatal/severe injuries to the riders should be identified by adopting accurate modeling techniques. This study develops statistical and machine learning models to predict injury severity of two-wheeler riders involved in crashes and thereby identify factors contributing to the crashes. Models with highest prediction accuracy are selected for identifying significant variables and their effects on likelihood of severe crashes. Most TW injury severity studies have been conducted to analyse the impacts of crashes on TW drivers. During a crash, both driver and pillion rider get equally exposed to the impact. Hence, it is important to explore the impacts of crashes on the pillion riders as well. The study analysed the causes of severe injuries to TW drivers and pillion riders separately.
The study further discusses a model framework for simultaneously predicting TW driver and pillion rider injury severity and improving prediction accuracy of model by accounting for spatial correlation in crash data. This advanced modeling approach incorporates potential correlation in injury severities of TW riders and spatial heterogeneities simultaneously. The results suggest that the incorporation of spatial dependence in crash data significantly improved the prediction accuracy of the model.