Traffic Data Analysis Using Kernel PCA and Self-Organizing Map

In intelligent transportation system, one of the most difficult tasks is to manage the mass amount of data and discover useful information from them, so data mining plays an important role in extracting temporal and spatial relations in a networked system. In this paper, we propose a novel method for traffic data analysis. Kernel principal component analysis (KPCA) is used to reduce data dimensionality and extract features from them, then self-organizing map (SOM) is applied in the unsupervised clustering of links. Subject interpretation and regression equations are used to analyze the clustering result. Case studies on real data from UTC-SCOOT System in Beijing prove that the proposed method is effective in extracting non-linear relations between different links and revealing hidden patterns in traffic flow data. The result yielded can support further analysis, like traffic parameter forecasting and traffic flow control