WIT Press


One-against-all Multicategory Classification Via Discrete Support Vector Machines

Price

Free (open access)

Paper DOI

10.2495/DATA030251

Volume

29

Pages

10

Published

2003

Size

445 kb

Author(s)

C. Orsenigo & C. Vercellis

Abstract

One-against-all multicategory classification via discrete support vector machines C. Orsenigo & C. Vercellis Dipartimento di Ingegneria Gestionale, Politecnico di Milano, Italy Abstract Discrete support vector machines (DSVM), recently proposed in [lo] and [ l l ] for binary classification problems, have been shown to outperform other competing approaches on well-known benchmark datasets. Here we address their extension to multicategory classification, by developing a one-against-all framework in which a set of binary discrimination problems are solved by means of DSVM. Computational tests on publicly available datasets are then conducted to compare multicategory DSVM with other methods, indicating that the proposed technique achieves high accuracies. 1 Introduction Classification problems with multiple categorical target classes play a central role in several applications of data mining in such diversified fields as marketing, finance, fraud detection, text and speech categorization, medical

Keywords