Detection of Outliers in an Unsupervised Environment

摘要: Outliers are exceptions when compared with the rest of data. do not have a clear distinction respect to regular samples in dataset. Analysis and knowledge extraction from data outliers lead ambiguity confused conclusions. Therefore, there is need for detection as pre-processing stage mining. In multidimensional perspective, outlier challenging issue an object may deviate one subspace appear perfectly another subspace. this paper, ensemble meta-algorithm has been proposed analyze vote identification subspaces. Cook’s distance, regression based model applied detect voted by meta-algorithm. Extensive experimentation on real datasets demonstrates efficiency system detecting outliers.