An Efficient Mesoscopic Modeling Method for Large Volume Traffic Flow Using Process Mining Techniques

Abstract

With the development of computing power and the widespread use of sensor technologies, highly accurate and frequent large-volume traffic flow data has become readily available. Model creation from these traffic flow data can be used for various purposes but handling large-volume traffic flow data requires huge computing power and a great deal of work. To mitigate this problem, we study mesoscopic models in which continuous values are replaced with statistical information derived from reduced data by discretization while retaining the model abstraction level that allows for bottleneck verification and identification of stagnation. In addition, we propose a novel model creation method that reduces the workload by applying process mining techniques. Furthermore, using airport traffic flow data as an example, we create an actual model and show that process mining techniques are quite useful in the modeling process.