Transport mode identification by clustering travel time data

Travel time data of road users collected by Bluetooth scanners are of great value in traffic monitoring and planning. To estimate the travel time of road users over a segment of road, discriminating between different types of travellers is essential, but often overlooked by researchers. This paper explores the feasibility of transport mode identification using clustering methods. The performance of the k-means clustering algorithm and the Gaussian mixture model is examined via an empirical study of travel time data collected from road segments in the north Brisbane region, Queensland, Australia. It is demonstrated that both clustering methods are able to detect multiple transport modes and produce travel time estimates that are close to reality. The methods and results provide a guideline for transport mode identification, and may contribute to further issues related to traffic monitoring such as forecasting and planning.