Downloads: 2 | Views: 96 | Weekly Hits: ⮙1 | Monthly Hits: ⮙1

Research Paper | Statistics | India | Volume 12 Issue 2, February 2023

Designing a better Support Vector Machines Classification Model for Multi-Class Category

Muthukrishnan .R | Udaya Prakash .N

Abstract: Support vector machines are most widely used classification technique and are computationally efficient. Even though, it provides better classification for high-dimensional data, it suffers a lot when the data contains extremes. Such situation kernel function plays a vital role to produce reliable results. An attempt has been made for selecting the best technique for a model from the multi-class groups in order to achieve the best accuracy level using Support Vector Machine (SVM) with a significant option among efficiency and predictiveness. Further, before scheming the best kernel, it has been tried to find the best type of SVM based on multiclass methods in order to select the superior one and save time in its application. When a poor type or method is used, it will result in a significant loss in its predictive accuracy, which will lead to high misclassification rate. In this context, this paper investigates the various types of SVMs and compares their accuracy levels by computing the misclassification rate on real and simulation environment. It shows that the multi-class SVM approach outperforms the others in terms of accuracy.

Keywords: Support Vector Machine, Multi-class classification, Kernels

Edition: Volume 12 Issue 2, February 2023,

Pages: 905 - 909