Survey of Improved k-means Clustering Algorithms: Improvements, Shortcomings and Scope for Further Enhancement and Scalability

Clustering algorithms are popular algorithms used in various fields of science and engineering and technologies. The k-means is example unsupervised clustering algorithm used in various applications such as medical images clustering, gene data clustering etc. There is huge research work done on basic k-means clustering algorithm for its enhancement. But researchers focused only on some of the limitations of k-means. This paper studied some of literatures on improved k-means algorithms, summarized their shortcomings and identified scope for further enhancement to make it more scalable and efficient for large data. From the literatures this paper studied distance, validity and stability measures, algorithms for initial centroids selection and algorithms to decide value of k. Then proposing objectives and guidelines for enhanced scalable clustering algorithm. Also suggesting method to avoid outliers using concept of semantic analysis and AI.