Based on traffic characteristic analysis, the highway managers could be aware of road capacity changes, and better traffic forecast or traffic management approaches could be achieved. For instance, the flow curve is a common index to reflect the highway operation in highway management and traffic flow forecasting. By using the flow curve classification method, it is possible to explore the traffic changes and then deliver scientific traffic management. Actually, clustering methods were widely applied in curve classification, but very few in traffic flow curve analysis. In this paper, a traffic flow curve classification method based on curve similarity and using the Density Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm was presented. Firstly, due to the discreteness of the sectional flow curve, the discrete Frechet distance was selected to measure the similarity of the flow curves. Next, the DBSCAN clustering algorithm was used to classify the flow curves into different categories. Then, the characteristics of each kind of flow curve were achieved and the change characteristics could be obtained. At last the method was applied in a real sectional flow data based flow curve classification and the results showed that the proposed curve similarity based flow curve classification method would be more accurate and efficient.

CICTP