Using multinomial regression to explore the spatial factors affecting left-turn oncoming accidents involving motorcycles

Abstract

Objective

In mixed traffic with a high proportion of motorcycles, accidents involving motorcycles usually account for the highest percentage of total accidents and are thus important. Because the motorcycle can easily weave around within the traffic flow, leading to traffic conflicts and accidents, this research analyzed left-turn oncoming collisions at intersections and explored the impacts of crossing positions of the motorcycles and opposite left-turn cars on accident risk. Furthermore, based on the analysis results, possible ways to prevent crashes are proposed.

Methods

We collected accident videos to develop an image dataset to explore the spatial factors affecting left-turn accidents by multinomial regression model (MNL). In addition, to collect samples without incidents as the reference category of the MNL, this research used post encroachment time analysis to screen the samples of general traffic flow. Finally, elasticity analysis was employed to explore the influences of the spatial factors on incident risk probability.

Results

The results showed that the accident risk was significantly affected by the cross-section locations of the left-turn vehicles and motorcycles during the passing process. intersection.

Conclusions

When a straight motorcycle travels in the curb lane or an opposite left-turn car does not turn with an appropriate turning trajectory, the risk of a left-turn oncoming collision will increase. Therefore, even if a straight motorcycle has the right-of-way, the operator should choose an appropriate cross-section position and avoid riding to the right of a car when passing through an intersection.

KEYWORDS:

• Accident prevention
• motorcycle
• multinomial logistic regression
• mixed traffic
• motorcyclist
• spatial factor

Additional information

Funding

This research was supported by the Ministry of Science and Technology in Taiwan.