Survey on Various Enhanced K-Means Algorithms

Abstract: Data Mining is defined as a technique used to extract and mine the invisible, meaningful information from mountain of data. Clustering is an important technique that has been introduced in the area of data mining. Clustering is defined as a method used to group similar data into a set of clusters based on some common characteristics. K-means is one of the popular partitional based clustering algorithms in the area of research. The impact factor of k-means is its simplicity, high efficiency and scalability. However, is also comprises of number of limitations: random selection of initial centroids, number of cluster K need to be initialized and influence by outliers. In view of these deficiencies, this paper presents a survey of improvements done to traditional k-means to handle such limitations.

Keywords: Data Mining, Clustering, K-means algorithm, Improved K-means algorithm