Ensemble Pruning Using Harmony Search

Abstract

In recent years, a number of works proposing the combination of multiple classifiers to produce a single classification have been reported. The resulting classifier, referred to as an ensemble classifier, is generally found to be more accurate than any of the individual classifiers making up the ensemble. In an ensemble of classifiers, it is hoped that each individual classifier will focus on different aspects of the data and error under different circumstances. By combining a set of so-called base classifiers, the deficiencies of each classifier may be compensated by the efficiency of the others. Ensemble pruning deals with the reduction of an ensemble of predictive models in order to improve its efficiency and performance. Ensemble pruning can be considered as an optimization problem. In our work we propose the use of Harmony search, a music inspired algorithm to prune and select the best combination of classifiers. The work is compared with AdaBoost and Bagging among other popular ensemble methods and our method is shown to perform better than the other methods. We have also compared our work with an ensemble pruning technique based on genetic algorithm and our model has shown better accuracy.