Development of Traffic Accident Prediction Models Using Traffic and Road Characteristics : A Case Study from Sri Lanka

Road accident is one of the major problems for most communities in developing countries, where 90% of the world road accident fatalities occur (WHO, 2009), which require serious attention in searching for preventive measures to minimize this problem. The traffic accident data in developing countries are just merely statistics which will not lead to further analyses and detailed studies. A section of highway of 20.5km in western province, Sri Lanka, was subdivided into approximately 200m segments and considered for this case study. Black spots were identified using three accident indices: (i) Accident Rate, (ii) Accident Frequency and (iii) Accident Severity. The study attempted to identify the relationship among Number of accidents (Y), Average Daily Traffic Flow in thousands (X1), Commercial Land Use Area in Square kilometers (X2), Binary variable to represent the vicinity of intersection (X3; if yes:=1, else:=0). The following relationship was found: Y=(-6.54)+0.23X1+0.94X2+12.89X3 where R-Sq=52.3%, R-Sq(adj)=51.0%, Std.err=11.61.