Mining related information of traffic flows on lanes by k-medoids

Nowadays, processing traffic flows has become an important part in intelligent transportation system (ITS). Prediction and estimation of flows, as a main application in this field, has gradually developed. Moreover, there exist some inherent relationships among various traffic flows, and the mining of related information can provide a platform for traffic flow prediction and estimation, and it can supply some guidance to layout traffic sensors. This paper presents a method of cluster by k-medoids to mine related information of traffic flows from spatial dimension. From spatial dimension, road lanes are clustered by k-medoids to constitute a table of related information. In order to make the mining of related information of flows more accurate, degree of saturation is also used to cluster related information. The results indicate that cluster through combination of flow and degree of saturation has a higher efficiency, and cluster by k-medoids outperforms that by k-means in all experiments.