Abstract:
The primary objective of this study is to investigate the gap acceptance behavior of vehicles at
unsignalized intersections. To achieve this, data were collected from three different
unsignalized T-intersections using video cameras, and a semiautomatic tool, Kinovea, was
utilized for data extraction. Two models were developed in this study, namely, the gap
acceptance decision model, which classifies the drivers' decision to accept or reject a gap based
on certain influencing factors, and the accepted spatial gap estimation model, which estimates
the gap accepted by drivers based on these influencing factors. In order to build these models,
various machine learning techniques such as logistic regression, support vector machine,
random forest and linear regression were employed for classification and regression analysis.
Furthermore, feature selection was applied to reduce the dimension of the data. The results
showed that the variable gap size had a positive relationship with gap acceptance behavior
while the variable speed had a negative relationship. Moreover, the random forest model
outperformed other models, achieving higher performance matrices. Conflicting speed was
found to be the most important variable in regression analysis, and other variables like waiting
time, number of queued vehicles, number of rejected gap, and minor road vehicle type also had
some importance in the model development. This study provides valuable insights into gap
acceptance behavior at unsignalized intersections and can be useful for traffic planning and
management.
