Reduction of ensemble of classifiers with a rule sets analysis

Abstract

The article shortly discusses the aim of classification task and its application to different domains of life. The idea of ensemble of classifiers is presented and some aspects of grouping methods are discussed. The paper points to the need of ensemble classifier pruning and presents a new approach for ensemble reduction. The proposed method is dedicated to committees of decision trees and bases on transformation of a tree set into a rule set and the new, suited to the pruning method, the weighted voting algorithm is also presented. There are also described experiments showing properties and effectiveness of the proposed method. Finally, directions of further research are mentioned.