Improving Support Vector Machine using Modified Kernel Function

Abstract

In machine learning, support vector machine (SVM) classifiers are highly effective in classification and pattern recognition tasks. The kernel function plays a crucial role in SVM, as it maps data into higher dimensional space to enable linear separability. This research proposes a modified kernel function, the Combined Polynomial Radial Basis Function (PRBF), that combines the strengths of Gaussian Radial Basis Function (RBF) and Polynomial kernels to improve the performance of SVM. The PRBF model aims to enhance classification accuracy by capturing patterns in the data at both local and global levels. The results demonstrate that the PRBF outperforms traditional RBF and Polynomial kernels across various datasets, achieving higher accuracy, sensitivity, and specificity. the PRBF kernel represents a significant advancement in SVM classifier technology, merging the advantages of RBF and Polynomial kernels to deliver enhanced accuracy, sensitivity, and specificity. This study validates that PRBF as a superior alternative for enhancing SVM performance in diverse applications in different domain.