International Journal of Advances in Computer Science and Its Applications
Author(s) : MD. MOSTAFIZER RAHMAN, MD. SOHRAB MAHMUD, MD.NASIM AKHTAR
Clustering is the process of grouping data into a set of disjoint classes called cluster. It is an effective technique used to classify collection of data into groups of related objects. K-means clustering algorithm is one of the most widely used clustering techniques. The main puzzle of K-means is initialization of centroids. Clustering performance of the K-means totally depends upon the correctness of the initial centroids. In general, K-means randomly selects initial centroids which often show in poor clustering results. This paper has proposed a new approach to optimizing the designation of initial centroids for K-means clustering. We propose a new approach for selecting initial centroids of K-means based on the weighted score of the dataset. According to our experimental results the new approach of Kmeans clustering algorithm reduces the total number of iterations, improve the time complexity and also it has the higher accuracy than the standard k-means clustering algorithm.