Investigations on performance of improved distinct cluster identification for two dimensional dataset

Abstract

In recent days, unsupervised clustering technique acts as a backbone for various research fields and applications such as pattern recognition system, data mining, big data, bio-informatics, machine learning, biomedical, biotechnology, web mining, image mining, sentimental analysis and image segmentation and so on. The unsupervised clustering technique is intended to identify dissimilar clusters in dataset based on user input for deeper investigation. It is classified into two different categories: Hierarchical and Partitioning. Agglomerative clustering scheme is a very good example of Hierarchical type and K-Means is a perfect example for Partitioning type. These schemes are fail to automatically identify appropriate number of dissimilar clusters in two dimensional dataset. The goal of the present newlineresearch has been to design various improved unsupervised clustering schemes such as Agglomerative and K-Means for robotically identifying suitable number of dissimilar clusters over large datasetswithout user input.An enhanced unsupervised clustering scheme, namely improved Limited Iteration Agglomerative Clustering (iLIAC) has been designed, to overcome the drawbacks in the existing agglomerative clustering technique. It aims to spontaneously separate the distinct clusters in dataset based on various optimal merge costs.Initially, it calculates the optimal merge costs newlineOMC and OMC+ over the dataset based on standard variance and standard deviation operations. newline newline