Framework for novel subspace clustering using search optimization methodology

Abstract

Improving the yield as well as the perform of subspace clustering is one of the less-investigated topics in high-dimensional data. After reviewing existing approaches, it seriously felt that there is a need for classification of data points retrieved from a different number of subspace. The proposed study has presented a novel framework that targets to improve the accuracy of subspace clustering by addressing the problem associated with the exist of occlusion noise and dimensional complexity. An analytical approach as been proposed to design this framework with more emphasis on outlier minimization followed by obtaining optimal clusters. The technique also introduces a simple search optimization method, which is less iterative and is more productive for identifying the élite outcomes in each iterative step. The study outcome shows superior accuracy with a low rate of error when compared with the conventional approach.