SimToll: A Highway Toll, Lane Selection, and Traffic Modeling Dataset

This paper presents a dataset about a toll highway consisting of a toll lane, a carpool lane, and three regular lanes. The dataset contains traffic information, like the number of vehicles on each lane type and the average speed on each lane, at intervals of 6 minutes. The dataset also provides information about the individual drivers/vehicles on the highway, like their departure and arrival times and the lane used. The dataset contains a total of 90 scenarios that cover varying the driver population size, the toll price, and the overall percentage of vehicles eligible to use the carpool lane. The simulations utilize a fuzzy logic engine to emulate the process of lane selection by drivers. In order to test and demonstrate the usefulness of the dataset, machine learning-based models were built to predict whether a driver would arrive late or not at his/her final destination based on his/her lane choice and the current road conditions. Four different classification algorithms were used to compare the performance. These models achieved accuracy above \(95\%\), with precision and recall metrics above \(90\%\).