Nonlinear System Identification Based on Multi-resolution Support Vector Regression

A novel support vector kernel, namely the multiresolution kernel, is presented based on the theory of multiresolution analysis of wavelet transform. The multi-resolution kernel is composed by the scaling function at some scale and wavelets with different resolution. Based on multi-resolution kernel and support vector machine, a new regression model that is called multi-resolution support vector regression (MRSVR) for function regression is constructed. The MR-SVR used to nonlinear system identification, not only has the advantages of support vector machine, but also has the capability of multiresolution which is useful to approximate nonlinear function. Simulation examples show the feasibility and effectiveness of the method