ResearchBib Share Your Research, Maximize Your Social Impacts
Sign for Notice Everyday Sign up >> Login

A COMPARATIVE STUDY ON K-MEDOIDS ALGORITHM WITH DENCLUE-IM APPROACH FOR BIG DATA

Journal: International Journal of Computer Science and Mobile Computing - IJCSMC (Vol.6, No. 11)

Publication Date:

Authors : ;

Page : 86-93

Keywords : cluster; clustering method; data mining; k-medoids; Big data; clustering;

Source : Downloadexternal Find it from : Google Scholarexternal

Abstract

Every day, a bulky volume of knowledge is generated by multiple sources namely social networks, mobile devices, Clustering plays a really very important role in exploring knowledge, making predictions and to nullify the anomalies within the knowledge. This type of knowledge sources turn out AN heterogeneous knowledge, that needs to be square measure engendered in high frequency. One among the techniques permitting to a stronger use and exploit this type of complicated knowledge is called a clump. Finding a compromise between performance and speed interval gifts a serious challenge to classify this humongous knowledge at our disposal. For this purpose, we have an attendency to propose AN economical algorithmic program that is studied in program that is AN improved version of DENCLUE, referred to as by the name DENCLUE-IM. The concept behind is to hurry calculation by avoiding the crucial step in DENCLUE that is that the Hill ascent step. Experimental results victimisation giant datasets proves the potency of our projected algorithmic program. Clusters that contain collateral, identical characteristics during a dataset square measure classified victimisation repetitive techniques. However, because of the increase in global data is growing day-by-day terribly increasing giant datasets with very little or no information are often known into attention grabbing patterns with clump. In this comparative study two most well liked clump algorithms K-Means and K-Medoids square measure evaluated on data set transaction 10k of KEEL. The input to those algorithms square measure arbitrarily distributed knowledge points and supported their similarity clusters has been generated. The comparison results show that point taken in cluster head choice and area quality of overlapping of cluster is way higher in K-Medoids than K-Means. Additionally K-Medoids is healthier in terms of execution time, non sensitive to outliers and reduces noise as compared to K -Means because it minimizes the total of dissimilarities of knowledge objects.

Last modified: 2017-12-05 14:58:51