LASSO regression to determine risk factors for road accident casualties in Malaysia in the presence of multicollinearity.

This paper studied the risk factors that affect road accident casualties in Malaysia as it has a positive impact on the death index of the country. A total of five risk factors including driver, vehicle, road, weather and pedestrian were studied in this paper. Three different LASSO models: without the removal of multicollinearity and influential observation (Case A), removal of multicollinearity (Case B) and influential observation (Case C) are formed to identify the significant risk factors that influence road accident casualties in Malaysia. Based on the findings, Case C with the smallest Mean Square Error of Prediction, MSE(P) is chosen as the best LASSO model in this paper. Therefore, variables in Case C including overloading passengers, drug, dangerous turning, driving too close, not conforming to traffic light, road shoulders low/high, potholes, slippery road, no guard rails, straight, roundabout, rethread, daylight, dawn/dusk, dark without street-light and lastly infirmity act as the significant risk factors that lead to an increment of road accident casualties. With these findings, some appropriate and strategic precautions can be implemented by the authorities to reduce road casualties in Malaysia. [ABSTRACT FROM AUTHOR]