Home // DBKDA 2011, The Third International Conference on Advances in Databases, Knowledge, and Data Applications // View article

New fuzzy multi-class methode to train SVM classifier

Authors:
Taoufik Guernine
Kacem Zeroual

Keywords: Classification, SVM, Fuzzy logic.

Abstract:
In this paper, we presente a new classification methode based on Support Vector Machine (SVM) to treat multi-class problems. In context of multi-class problems, we have to separate large number of classes. SVM becomes an important machine learning tool to handle multi-classe problems. Usually, SVM classifier is implemented to deal with binary classification problems. In order to handle multi-class problems, we present a new method that builds dynamically a hierarchal structur from training data. Our multi-class method is based on three main concepts: Hierarchical classification, Fuzzy logic and SVM. We combine multiple binary SVMs to solve multi-class problems. The proposed method divides the original problem into su-problems in order to reduce tis complexity.

Pages: 77 to 82

Copyright: Copyright (c) IARIA, 2011

Publication date: January 23, 2011

Published in: conference

ISSN: 2308-4332

ISBN: 978-1-61208-115-1

Location: St. Maarten, The Netherlands Antilles

Dates: from January 23, 2011 to January 28, 2011