The impact of real-time traffic parameters on crash severity has only recently been explored. The current research adds to knowledge by examining data mining techniques to investigate the determinants behind injury severity of occupants involved in crashes in the urban motorway Attica Tollway Athens, Greece. The proposed methodological approach involves a three-level process. Firstly, a two-step cluster analysis is performed to classify data into different groups (clusters). Secondly, a factor analysis is applied to group a number of relevant traffic variables into appropriate factors. Lastly, binary logit regression models are used to correlate clusters and factors with crash injury severity. The required crash data were extracted from the database SANTRA of National Technical University of Athens, consisting of 387 injury cases, for the period 2006 to 2011. The findings of this study demonstrate that hazardous traffic conditions are identified and real-time safety management policies can be promoted and improved.
|Tags||accident analysis, big data, machine learning, motorways, statistical modelling|