Identifying the key risk factors of traffic accident injury severity on Slovenian roads using a non-parametric classification tree

Abstract

From both a practical and economic point of view, road transport meets almost all the requirements of modern life, but it is also a source of numerous negative effects, including traffic accidents. In order to design a safe transport system and achieve the ‘zero vision’ goal – no serious injuries or fatalities in traffic accidents – there is a growing need for a systematic approach to this problem. Prior to the assessment of any accident prevention measure it is necessary to identify the most important factors and significant patterns which affect the severity of accidents and injuries. In this study, the crash data from Slovenia pertaining to the period 2005–2009 were analysed with a Classification and Regression Tree (CART) algorithm, one of the most widely applied data mining technique when analysing a large amount of data with several independent quantitative or qualitative variables. Before building a non-parametric classification tree, the data were split into three totally separate subsets, the training set, the testing set, and the evaluation set. Moreover, using the Variable Importance Measure (VIM) the factor of influence of nine independent variables on the target variables were calculated. The results confirm that traffic accidents and injuries on Slovenian roads are caused by a combination of factors, the most important of them being human error, or more precisely, speeding and driving in the wrong lane.

Keywords:

• road traffic accident
• injury severity
• variable importance
• classification error
• classification tree