Archives

Flexible K-Medoids Clustering Technique for Big Data Analytics


C.K. Praseeda and Dr.B.L. Shivakumar
Abstract

Now a day’s data is generating at an exponential rate by machines and human interactions with the development of internet technologies and electronic devices. The enormous amount of data being generated is stored and analyzed in different formats. Each data processed from the different scenario especially in the field of the telecom data needs more analytical development and validation for better processing outcomes. Data Mining refers to a process of extraction of earlier unknown and potentially resourceful knowledge from the massive amount of data. However, big data analytics support to improve the growth and the productivity of the industry. The selection of a good data mining technique to obtain the accurate result on a particular dataset is very important. The goal of this paper is to define a suitable clustering technique to analyze the generated clusters and determine the appropriate patterns based on the customers’ behavior using telecom data. In this research work, a Flexible K-Medoids algorithm is developed that sorts and consolidates the hybrid enhancement of K-Means, K-Median and K-Medoids algorithms which gives a deep analytics on the telecom data that reflects in the outcome of the strategic consolidation of the customers’ behaviour. This paper also analyses the efficiency of proposed hybrid algorithms for discovering clustering rules between items in a large database to produce maximum accuracy as possible with less number of iterations to converge. The comparison towards the performance is analyzed under various criteria.

Volume 11 | 08-Special Issue

Pages: 1455-1462