期刊名称：BVICAM's International Journal of Information Technology
印刷版ISSN：0973-5658
出版年度：2013
卷号：5
期号：1
语种：English
出版社：Bharati Vidyapeeth's Institute of Computer Applications and Management
摘要：In this paper, a recent yet powerful technique for classification of datasets is presented. The paper contributes to highlight the importance of an ensemble approach over individual classifiers to achieve better classification accuracy of a classifier. In this paper, given dataset is divided into a number of parts to constitute an ensemble. The ensemble combines these classifiers. An unknown data pattern is tested on the ensemble. Using bagging, majority of voting technique, the performance of ensemble is determined on different sections of datasets. In the paper, six bench mark datasets are used for investigation. Each dataset is trained with 80%, 60% and 50% of the data patterns for classification. The number of classifiers in an ensemble for each data set is changed to 5,7 and 9. As a typical case, k?nearest neighbor (k-NN) classifiers are used with the values of k varying to 1,3 and 5. The classification accuracies of individual classifiers and those of ensembles are computed at each case. After extensive experiments of proposed scheme, by taking random shuffling and selection of data patterns for training and testing, it is observed that in every case, the classification accuracy obtained by ensemble is higher than that obtained by individual classifier.
关键词：Index Terms - Classification;Ensemble of classifies; bagging;k-nn classifier.
