Academic Journals Database
Disseminating quality controlled scientific knowledge

Performance Evaluation of Kernels in Multiclass Support Vector Machines

ADD TO MY LIST
 
Author(s): R. Sangeetha | B. Kalpana

Journal: International Journal of Soft Computing & Engineering
ISSN 2231-2307

Volume: 1;
Issue: 5;
Start page: 138;
Date: 2011;
VIEW PDF   PDF DOWNLOAD PDF   Download PDF Original page

Keywords: Support Vector Machine | Multiclass Classification | Kernel function | One versus One | One versus All.

ABSTRACT
In recent years, Kernel based learning algorithmhas been receiving increasing attention in the research domain.Kernel based learning algorithms are related internally with thekernel functions as a key factor. Support Vector Machines aregaining popularity because of their promising performance inclassification and prediction. The success of SVM lies in suitablekernel design and selection of its parameters. SVM istheoretically well-defined and exhibits good generalizationresult for many real world problems. SVM is extended frombinary classification to multiclass classification since manyreal-life datasets involve multiclass data. In this paper, wepropose an optimal kernel for one-versus-one (OAO) andone-versus-all (OAA) multiclass support vector machines. Theperformance of the OAO and OAA are evaluated using themetrics like accuracy, support vectors, support vectorpercentage, classification error, and speed. The empirical resultsdemonstrate the ability to use more generalized kernel functionsand it goes to prove that the polynomial kernel’s performance isconsistently better than other kernels in SVM for these datasets.
Save time & money - Smart Internet Solutions     

Tango Rapperswil
Tango Rapperswil