A new trigonometric kernel function for SVM

In recent years, several machine learning algorithms have been proposed. Among of them, kernel approaches have been considered as a powerful tool for classification. Using an appropriate kernel function can significantly improve the accuracy of the classification. The main goal of this paper is to introduce a new trigonometric kernel function containing one parameter for the machine learning algorithms. Using simple mathematical tools, several useful properties of the proposed kernel function are presented. We also conduct an empirical evaluation on the kernel-SVM and kernel-SVR methods and demonstrate its strong performance compared to other kernel functions.