Nonlinear feature selection based on hybrid KCCA-FNN algorithm for modeling

Abstract

A hybrid algorithm based on kernel canonical correlation analysis (KCCA) and false nearest neighbor method (FNN) for selecting variables to reduce redundant feature and increate accuracy in nonlinear system modeling. In the proposed method, the KCCA can be employed to overcome difficulties encountered with the existing multicollinearity between the factors, the FNN can be used to calculate the variables' map distance in the new KCCA feature space to select secondary variables. Comparing with the fully parametric model, the method is provided for the variable selection of nonlinear system modeling for the production processing of hydrogen cyanide.