Home Author Index Search Volume 1 May 2009 ISSN 1797-9617

International Journal of

Recent Trends in Engineering

Home > Vol. 1, No. 2

 

International Journal of Recent Trends in Engineering (IJRTE)

ISSN 1797-9617

Volume 1, Number 2, May 2009

Issue on Computer Science

Page(s): 23-25

Comparative study of Linear and Quadratic Programming Versions of SVM on Various Real Life Data Sets

Hemalatha.T, Sukumar.K, Mithila.T, Soman.K.P

Full text: PDF

Abstract

In this paper we are comparing the performance of Linear Programming and Quadratic Programming versions of SVM on real life data sets and most of the real life data sets are not linearly separable. Quadratic programs are more complex and are usually not taught in Undergraduate studies, while linear programming is easy to understand and is available even in spreadsheet packages like Microsoft Excel. With undergraduate students in mind we did a comparative study to ascertain the accuracy differences. Quadratic version of SVM is reformulated as linear formulation by avoiding the quadratic term in Quadratic formulation and the results of both Quadratic formulation and linear formulation are compared. Percentage of accuracy is assessed by performing 10 fold cross validation on the data. It is found that linear programming is though expected to be less accurate; the accuracy difference is not too high.

Index Terms

 

Published by Academy Publisher in cooperation with the ACEEE

@ Copyright 2009 ACADEMY PUBLISHER All rights reserved