Research on prediction of protein sub-cellular location based on KLDA with combined kernel function

Abstract

To improve the accuracy of prediction of sub-cellular location, a new method using kernel linear discriminant analysis with combinational kernel function which is made up of the Gauss kernel function and the polynomial kernel function is used to the predict the sub-cellular location. In order to confirm the reliability of the research, the data used in this paper are from the standard data set included in Swiss-Prot database and the values of parameters for combined kernel function are determined reasonably. The results indicate that the proposed method with combined kernel function is more efficient than the kernel linear discriminant analysis algorithm with traditional kernel functions in the prediction of sub-cellular location.