An Empirical Analysis of Multiclass Classification Techniques in Data Mining

Abstract

Data mining has been an active area of research for the past couple of decades. Classification is an important data mining technique that consists of assigning a data instance to one of the several predefined categories. Various successful methods have been suggested and tested to solve the problem in the binary classification case. However, the multiclass classification has been attempted by only few researchers. The objective of this paper is to investigate various techniques for solving the multiclass classification problem. Three nonevolutionary and one evolutionary algorithm are compared on four datasets. Further, using this analysis, the paper presents the benefits of producing a hybrid classifier by combining evolutionary and non-evolutionary algorithms; specifically, by merging Genetic Programming and Decision Tree.