The Research on Email Classification Based on Q-gaussian Kernel Svm

The use of different kernel functions in SVM (Support Vector machines) has been reported in the literature. In this paper, the use of the q-Gaussian function as kernel function in SVM is investigated and q-Gaussian function is explored. While the q-Gaussian kernel SVM classifiers being built, cross validation methods are used to select the non-extensive entropic index q under varying feature sizes, the punishment parameter and kernel width are set empirically, the email classification on two leading Chinese email corpuses, TREC06c and CCERT, are implemented with SVM classifiers employing Gaussian kernel and q-Gaussian kernel. Experiment results show that q-Gaussian kernel SVMs can enhance the classification performance effectively.