Data mining is used to extract the hidden patterns from large datasets and extracted patterns are helpful to identify knowledge about data to users. As various approaches are there for data mining named Clustering, Classification, Association rule mining, etc. Amongst all we consider clustering, which is an unsupervised learning and grouping. This paper demonstrates clustering technique named k-means clustering and its various improvements in different domains exterminate the limitations of traditional k-means clustering. K-means clustering is the simple partitioning clustering algorithm and exhibit many limitations, so it is very important to understand various enhancements for constructing hybrid algorithms to improve accuracy of algorithms. Various areas are defined where k-means clustering is widely used nowadays such as in healthcare, improving academic performance and optimization of search engine and much more.