This paper intends to provide a broad picture of traffic accidents in Israel by uncovering their patterns and determinants in order to answer an increasing need of designing preventive measures, addressing particular situations and targeting specific social groups with the ultimate objective of reducing the number of traffic fatalities and accidents. The analysis focuses on 1,793 fatal accidents occurred during the four-year period between 2003 and 2006, and applies data mining techniques with the objective of extracting from the data relevant information about accident patterns and major factors without a priori assumptions about the expected outcome of the study. Kohonen neural networks reveal five accident patterns: (i) single-vehicle accidents of young drivers; (ii) multiple-vehicle accidents between young drivers; (iii) accidents involving either motorcycles or bicycles; (iv) accidents where elderly pedestrians crossed in urban areas; (v) accidents where mostly young children and teenagers cross roads in small villages. Feed-forward back-propagation neural networks indicate that demographic characteristics of both victims and drivers are the most relevant determinants, and other significant factors are the road conditions, the accident location in either urban or rural areas, the accident location in either sections or intersections, and the period of the day when the crash occurs.
|Title of host publication||Proceedings of the 12th WCTR Conference|
|Publication status||Published - 2010|
|Event||12th World Conference on Transportation Research - Lisbon, Portugal|
Duration: 11 Jul 2010 → 15 Jul 2010
Conference number: 12
|Conference||12th World Conference on Transportation Research|
|Period||11/07/2010 → 15/07/2010|