Development of Driver Behavior’s Accident Prediction Models

Abstract

Amman, the capital city of Jordan, has been expanding in terms of size, investment, and growth of vehicles. Such a growth has lead to traffic
jams and delays experienced at all levels of services, and higher accident frequency level at several locations, which resulted in loss of people’s
lives, and causing major economical and social concerns in the country. Statistical models were employed to analyze accident frequency for
Amman. The objectives were to identify hazardous locations by developing accident prediction models. Accident data was collected and linked to
independent variables. Several models were developed to identify the relationship between accident frequency and key behavioral characteristics
of drivers. Different types of high-accident locations were identified, classified, and ranked according to their hazardous degrees by using statistical
techniques. Findings indicated that the short distance between vehicles, lane changing, and non-yielding right-of-way variables were the most
critical causes of accidents. A priority ranking for countermeasures was recommended to reduce accidents and improve the overall driving safety
at hazardous locations based on the developed models. Recommendations were made for the way in which accidents on these locations would be
treated. Suggestions were made for the practical and theoretical development for further research