Improved kernel learning using smoothing parameter based linear kernel

Abstract

Kernel based learning has found wide applications in several data mining problems. In this paper, we propose a modified classical linear kernel using an automatic smoothing parameter (Sp) selection compared with the existing approach. We designed the Sp values using the Eigen values computed from the dataset. Experiment results using some classification related benchmark datasets reveal that the improved linear kernel method performed better than some of the existing kernel techniques. © Springer-Verlag Berlin Heidelberg 2003.