Assessing Behavioral Patterns of Motorcyclists Based on Traffic Control Device at City Intersections by Classification Tree Algorithm

Abstract

According to the forensic statistics, in Iran, 26 percent of those killed in traffic accidents are motorcyclists in recent years. Thus, it is necessary to investigate the causes of motorcycle accidents because of the high number of motorcyclist casualties. Motorcyclists' dangerous behaviors are among the causes of events that are discussed in this study. Traffic signs have the important role of traffic controller, and road surface marking is a tool for traffic separation and has a significant effect on drivers' behaviors. The aim of this study is to investigate the effect of variables, including traffic conditions, motorcyclists' psychological conditions, and symptoms and function of traffic lights on the motorcyclists' dangerous behaviors. In this study, classification tree method is used to determine the effective factors in some motorcyclists' dangerous behaviors such as the amount of deviation from the center lane, lane changing, and running red lights. The classification tree is easy to understand and interpret because of the graphical display of results. The data classification tree is made based on the classification and regression tree algorithm (CRT) in this study. The data are collected from the 7 intersections in a city with the medium population by video-based observation method. Hand-held cameras randomly record the motorcyclists' motions and, then, these behaviors are investigated in the office by playing back the videos at slow motion. The obtained trees show that the variables of traffic volume have the greatest impact on the motorcyclists' diversion from the center lane and lane changing. Also, the clarity of the pavement marking is effective in reducing deviation from the middle lane by cyclists so that, in the streets with the line color contrast of more than 1.36, deviation from the center lane is reduced by 25 cm.