Performance Analysis of Standard K-Means and Enhanced k-Means Clustering Algorithm

Clustering is one of the most significant techniques of data mining to uncover the hidden relationship, discover patterns from the large complex datasets. This article discusses different approaches of clustering, different challenges of conventional k-means clustering like it produce different outputs depending upon the different initialization of centroids and how it calculates the distance of each object with all centroids. The enhancements to the K-means algorithm offer several benefits, including improved clustering accuracy, increased scalability, and more flexible distance measures. In the proposed work, enhanced k-means algorithm is designed to initialize the centroids systematically by percentile method and an array is used to store the distance of object to centroids in previous iteration and used in new iteration to reduce the time complexity of algorithm. The outcome of proposed methodology showed that the improved k-means algorithm outperforms the conventional k-means algorithm in terms of accuracy and efficiency.