Multicore Implementation of K-Means Clustering Algorithm

Abstract

Multi-core processing is extensively used in every sector for its performance efficiency, with the advent of multi-core architecture have to modify the existing primitive algorithms. This study analyses the feasibility of K-mean data-mining technique, which is applied to a hybrid cluster with multi-core programming. The algorithm is developed using Message Passing Interface (MPI) and C programming languages for the parallel processing of the sets and uses the CPU to its maximum power for the hybrid sets. The heterogeneous clusters are confirmed by the usage of MPICH2 (High performance and portability implementation of MPI). examined the algorithm for the huge dataset. The dataset is split into a number of cores and each of the cores estimates the number of dusters on the same dataset interdependent to each other. By this, assert the core processor time for communication is significant for huge datasets. Hence, the same dataset for two different processors takes different times even with identical speed and memory and also with different speeds and access times.