In the Proposed work, firstly we are discussing the various techniques of clustering and most widely used k-means clustering Algorithm and then design a hybrid technique from various enhancements done on k-means Algorithm that improves the performance of the system. Clustering techniques are used to implement unsupervised learning. In clustering K-means algorithm with various advancements is used for making clusters in large datasets. There are some limitations in basic k-means Algorithm like choosing of number of clusters initially before making the clusters and finding the initial centroids. In this work elbow technique using sum of square errors (SSE) is used for calculating number of clusters and Back Propagation Algorithm is used to do the iterations to check the Euclidian distance of different observations with centroids. By using this strategy, performance of the system is increased in terms of reducing the number of iterations for large datasets and all the data sets are comprised properly in different clusters. After studying all enhancements of k-means Algorithm, a hybrid Algorithm is designed that works efficiently for large datasets and scaling is used so that the unclustered datasets should also be included in clusters.