A REVIEW PAPER ON IMPROVED K-MEANS TECHNIQUE FOR OUTLIER DETECTION IN HIGH DIMENSIONAL DATASET

In many data mining application domain outlier detection is an important task, it can be regard as a binary asymmetric or unbalanced classification of pattern where one class has higher cardinality than the other, finding outlier is very challenging in high dimensional dataset where data contain large amount of noise which causes effectiveness problem, they are more useful based on their diagnosis of data characteristics which deviate significantly from average, this paper present Improved K-Means Technique for Outlier Detection in High Dimensional Dataset. Various subspace based method has been proposed for searching abnormal saprse density unit in subspace, this paper proposes a Clique density based clustering algorithm that attempt to deal with subspace that create dense reason when projected onto lower subspace in high dimensional data set and then apply the improved K-Means algorithm on generated subspaces for effectively and efficiently identifying outliers for getting the more meaningful and interpretable result.