Effective K-means clustering algorithm for efficient data mining

With the advancement of technology and the increase of data, Data mining needs to be more efficient. Therefore, data mining along with machine learning and statistics gets effective to face today's scenario. Machine learning provides an easy way to accomplish tasks and statics provides a standard computational skill to understand, manage and analyze data. The work surrounds the enhancement of the K-means clustering algorithm to make it more effective and to create accurate clusters over large data sets. There are some flaws in the classical K-means algorithm the major is that it cannot work over large and dense data points. This paper aims to improve the k-means algorithm so that it can create more smooth and more accurate clusters over highly dense data points. To accomplish this window density method is being used to create more homogeneous clusters. This will make the K-means algorithm suitable for all shapes and sizes of clusters. This will also enable handling noise effects, spotting outliers, and discovering clusters other than hyper ellipsoids. We have introduced the K-means algorithm with window density and derived that the results are more accurate than the original K-means algorithm as we have evaluated the outcomes through purity measure.