Enhanced K-Means++ Clustering For Big Data with Mapreduce

Abstract

Clustering big data using data mining algorithms is a modern approach, used in various science and medical fields. k-means clustering algorithm is a good approach for clustering, but choosing initial centers and provides less accuracy guarantees. The enhanced k-means approach called 𝑘-means++ chooses one center uniformly at random provides better functionality, but fails to handle data of larger volume in distributed environment. The mapreduce 𝑘-means++ method handles k-means++ algorithm by enhancing it in mapper and reducer phases, also reduces the no of iterations required to obtain 𝑘 centers. in which the 𝑘-means++ initialization algorithm is executed in the mapper phase and the weighted 𝑘-means++ initialization algorithm is run in the reducer phase. it reduces huge amount of communication and i/o costs. the proposed mapreduce 𝑘-means++ method obtains (𝛼2) approximation to the optimal solution of 𝑘-means.