COMBINATION OF DIFFERENCE SUBSPACE AND OPPURTUNISTIC CLUSTERING ON HIGH DIMENSIONAL DATA

Abstract

Clustering is recognized as sigificant technique for analysing data and concentric effort has been taken in different domains comprises of recognition of pattern, statistical analysis and data mining for decades. Subspace clustering is developed from the group of cluster objects from all subspaces of a dataset. During clustering of objects involing higher dimension, the accuracy and effectiveness of traditional clustering algorithms are very poor, because data objects may present in different clusters involving different subspaces of differing level of dimensions. To address the above issue, a new technique termed Difference Subspace and Opportunistic Clustering (DSOC) model is presented for high dimensional data to improve the accuracy during the search process and also considers the problem of accuracy in clustering the high dimensional data. Methods for obtaining subspace and designing clustering model for DSOC are specified and demonstrated, where the subspace identify the possibility of each cluster center with the detection of attackers based on multiple locations and estimation points with derived centroid points. Through comprehensive mathematical analysis, we show that DSOC improves the accuracy in high dimensional data with efficient cluster validation obtained using different subspace and opportunistic algorithm. As validated by extensive experiments on CORTINA and Ski Resort Data Set datasets DSOC produces high quality clusters by detecting the attackers and the efficiency of DSOC outperforms previous works.