Abstract

Support Vector Machine (SVMs) are renowned for their excellent performance in solving data-mining problems such as classification, regression and feature selection. In the field of statistical classification, SVMs classify data points into different groups based on finding the hyperplane that maximizes the margin between the two classes. SVMs can also use kernel functions to map the data into a higher dimensional space in case a hyperplane cannot be used to do the separation linearly. Using specific kernels allows us to model a particular feature space, and a suitable kernel can improve the SVMs' performance to classify data more accurately. We present a method to combine existing kernels in order to produce a new kernel which improves the accuracy of the classification and reduce the process time. We will discuss the theoretical and computational issues on SVMs. We are going to implement our method on a simulated data-set to see how it works, and then we will apply it to some large real-world data-sets.