Improved K-Means Clustering Algorithm Based on KD-Tree Approach

The Cluster Analysis is a vast area of application such as security, Image recognition, scientific investigation, business intelligence, biology, and web search. K-Means clustering algorithm is not performing well with huge data sets in terms of Execution time. To overcome this, A Parallel Approach is used to implement the K-Means algorithm using OpenMP API with the KD-Tree approach to provide dynamic load balancing, optimized execution time, and maintaining accuracy. The experiments are performed on handwritten digits and Bagofword data sets by using a system with multi-core. After the analysis of the Sequential approach and Parallel approach of implementation of K-Means, it is observed that the parallel approach outperforms with similar accuracy utilizing the computing resources available with the multi-core systems.