Method of Flow Curve Classification Based on Curve Similarity

Highway managers could use traffic characteristic analysis to understand capacity changes and forecast and manage traffic better. The flow curve is a common index to reflect highway operation in management and flow forecasting. Using flow curve classification, one can explore changes and deliver scientific traffic management. Clustering is widely applied in curve classification, but rarely in traffic flow curve analysis. This paper presents a traffic flow curve classification method based on curve similarity using density based spatial clustering of applications with noise (DBSCAN) algorithm. Due to sectional flow curve discreteness, discrete Frechet distance was selected to measure flow curve similarity. DBSCAN clustering algorithm was used to classify flow curves into different categories. Characteristics of each flow curve type were achieved and change characteristics were obtained. The method was applied in a real sectional flow data-based flow curve classification and results showed the proposed curve similarity-based flow curve classification method would be more accurate and efficient.