Development of a binary logistic lane change model and its validation using empirical freeway data

An effective macroscopic lane change (LC) model is required to facilitate active and dynamic lane management to develop cell-based multilane macroscopic traffic models. Existing logistic regression LC models do not undertake model classification of lane change; do not consider performance measures in the validation of field data and ignore movement between lanes during time-varying traffic. Models that consider the direction of LC are, however, biased in their prediction of left LC (LLC) and right LC (RLC) direction. This study proposed a simplified macroscopic cell-based binary logistic LC (BLLC) model describing the LC probability using fewer explanatory variables; in this model, the direction of LC is considered as a new variable. Considering the model performance measures, the results show that there exists substantial difference in LC behaviour in both directions. The present model also achieved a smaller difference in the percentage of accurate prediction (0.9%) between the LLC and RLC.