Using feature selection in identifying critical factors of injury severity

Prior literature often arbitrarily selects determinants of accident injuries to predict injury severity and thus ignores the usefulness of the chosen variables for effective prediction of injury severity. Furthermore, it is inappropriate to adopt the determinants of traff ic accidents from a particular country to examine another country's traff ic accidents. Taking the above into account, this paper applies the feature selection technique to identify critical factors of intoxicated driving to cause the injury severity in Taiwan, namely death and injury. Twenty five factors are chosen to perform feature selection. By the method, fifteen factors are classified into important factors, two marginal factors, and three unimportant factors