Drifting model approach to modeling based on weighted support vector machines

This paper proposes a novel drifting modeling (DM) method. Briefly, we first employ an improved SVMs algorithm named weighted support vector machines (W. SVMs), which is suitable for locally learning, and then the DM method using the algorithm is proposed. By applying the proposed modeling method to Fluidized Catalytic Cracking Unit (FCCU), the simulation results show that the property of this proposed approach is superior to global modeling method based on standard SVMs.