Mining for Similarities in Urban Traffic Flow Using Wavelets

To explore the temporal-spatial similarities existing in urban transportation network, a novel mining method is proposed to cluster the traffic flow series on different road links. Particularly, discrete wavelet transform is adopted for flow feature extraction, because it is insensitive to disturbance/scaling, and zooms in multiple finer granularities. Self-organizing map (SOM) algorithm is then used to cluster road links into groups, for which different feature sets are considered for different purposes. Two applications of outlier detection and traffic flow prediction are given as examples of the benefit of similarity mining. The proposed method is tested with the data collected from a typical urban traffic network, Beijing, China. Experiments indicate that the proposed method offers a more flexible analysis of multi-level similarities, which might be ignored by traditional methods.