Application of machine learning to characterize uneconomical managed lane choice behaviour

Abstract Models of tolled Managed Lane (ML) travel assume that people pay the toll to experience a shorter travel time than using the toll-free general-purpose lanes. However, there are times users pay to travel on the MLs but go slower than the toll-free lanes. This research examined these ‘uneconomical trips’ on MLs to discover potential similarities among these trips and help understand the unexpected lane choice behavior. Some potential factors considered were toll rate, traffic flow, and past trip experience. Random forest and logistic regression methods were proposed to study the impact and significance of variables on the probability of a user making an uneconomical ML trip. The results indicated toll rate, traffic flow, travel time variability, and trip route (start and end points) are the crucial elements in predicting uneconomical ML trips. This study is the first to examine uneconomical ML trips and attempt to understand these trips. This provides the first step towards being able to predict these trips and provides additional understanding of travel on MLs. Also, the findings of this study are the first step towards being able to include uneconomical ML trips into lane choice modeling and toll feasibility studies. Finally, this study proposed an unconventional tool (random forest) to predict the travel behavior on MLs and showed that such tools can perform better than currently used logistic regression models in predicting these uneconomical ML trips.