ResearchBib Share Your Research, Maximize Your Social Impacts
Sign for Notice Everyday Sign up >> Login

Multiclass Classification: A Review?

Journal: International Journal of Computer Science and Mobile Computing - IJCSMC (Vol.3, No. 4)

Publication Date:

Authors : ;

Page : 65-69

Keywords : ;

Source : Downloadexternal Find it from : Google Scholarexternal


Support Vector Machines (SVMs) are with success applied to resolve an outsized variety of classification and regression issues. SVMs were at first developed to perform binary classification; though, applications of binary classification are terribly restricted. The way to effectively extend it for multiclass classification remains Associate in nursing current analysis issue. Most of the sensible applications involve multiclass classification, particularly in remote sensing land cowl classification. Many ways are planned wherever usually we tend to construct a multiclass classifier by combining many binary classifiers. Because it is computationally costlier to resolve multiclass issues, comparisons of those ways victimization largescale issues haven't been seriously conducted. This paper compares the performance of six multi-class approaches to resolve classification drawback with remote sensing information in term of classification accuracy and machine price. we tend to then compare their performance with four ways supported binary classifications: “one-against-all,” “one-against-one,” and directed acyclic graph SVM (DAGSVM) and Error Corrected Output secret writing (ECOC) and combines their results to work out the category label of a take a look at pixel. Results from this study conclude that the “one-against-one” and DAG ways are additional appropriate for sensible use than the opposite ways. And conjointly indicate that for finding massive issues ways by considering all information quickly generally want less variety of support vectors.

Last modified: 2014-04-05 14:40:19