Analysis of Driver Injury Severity in Metropolitan Roads of India Through Classification Tree

Abstract

Reducing the injury severity from traffic accidents is most important step in mitigating accidents occurring in developing economies like India where two way roads are more common in cities. The number of deaths due to accidents has rose from 83,491 in 2005 to 1,36,071 in 2016 as per the latest reports of ministry of road transport and highways, government of India (MoRTH, 2016). To explore the factors contributing to injury severity in such roads, non parametric classification tree is used since it does not assume any underlying assumption between target variable and the predictors. CART (Classification and Regression tree), a classification tree establishes empirical relation between injury severity outcomes and variables including driver, vehicle, crash and environmental factors. The present study analyzed traffic crash data of single lane two way roads of Chennai city pertaining to period from January 2015 to December 2016. The final dataset included a total of 5271 crash information after excluding incomplete and missing data. This finalized dataset was split into two subsets, training and testing data and the classification models reported an accuracies of 63.4% and 61.5% for the training and testing data. The results indicated that collision type and vehicle type were the two important variables affecting the severity of injury. The findings of this study will help in determining influential factors so that countermeasures to reduce the severity of injury in urban cities can be developed.