A Modified DBSCAN clustering algorithm for the identification of PWA systems

This paper deals with the problem of identifying hybrid systems of class PWARX (PriceWise Auto-Regressive with Exogenous Inputs) based on the classification technique. This approach is insured by three main steps namely; data clustering, parameters vector estimation and regions estimation. This work is devoted to the data clustering step based on the DBSCAN (Density Based Spatial Clustering of Applications with Noise) clustering technique which has the ability to eliminate outliers and can group classes with arbitrary shapes. However, the performance of the DBSCAN algorithm degrades with dense adjacent clusters. Therefore, we have proposed an improvement of this algorithm. The simulation and experimental results show the efficiency and performance of the proposed approach.