Chapter 10 : Introduction to Scientific Data Mining : Direct Kernel Methods & Applications

The purpose of this chapter is to give a brief overview of data mining and to introduce direct kernel methods as a general-purpose and powerful data mining tool for predictive modeling, feature selection and visualization. Direct kernel methods are a generalized methodology to convert linear modeling tools into nonlinear regression models by applying the kernel transformation as a data pre-processing step. We will illustrate direct kernel methods for ridge regression and the self-organizing map and apply these methods to some challenging scientific data mining problems. Direct kernel methods are introduced in this chapter because they transpire the powerful nonlinear modeling power of support vector machines in a straightforward manner to more traditional regression and classification algorithms. An additional advantage of direct kernel methods is that only linear algebra is required.