Aspect Range Motivated Classifier Ensemble Reduction

This paper proposed about classifier which ensemble in data mining and it constitute one of the directions in main research of machine learning. In general multiple classifiers allow better predictable performance. To construct and aggregate ensembles there are several approaches that exists in the literature. To produce better results and to increase group diversity redundant members should be removed which contain in the ensemble systems. Smaller ensembles helps to relax storage requirements and the memory which improves efficiency by reducing run time overhead of the system. In this paper the ideas are extended for development of feature selection problems by transformation of training samples from ensemble predictions which supports reduction of classifier ensembles. To maximize the evaluation of feature subset from the selection of reduced subset of artificial feature this project use global heuristic harmony search. The large sized and high dimensional benchmark datasets are used to evaluate the resulting technique systematically it shows superior performance of classifications against randomly formed subsets and unreduced, original ensembles.