Implementation of Modified K-means Approach for Privacy Preserving in Data Mining

Abstract

Recent concerns regarding privacy breach issues have motivated the development of data mining methods, which preserve the privacy of individual data item. A cluster is gathering of information in such a way that the objects with similar properties are grouped into similar clusters and objects with dissimilar properties are placed into different clusters. The K-Means clustering algorithm is a broadly utilized plan to solve the clustering problem. In this paper, a comparative study of three clustering algorithms—K-means, Hierarchical and Cobweb across two different datasets is being performed. To form Clusters WEKA API has been used. The comparison is made with the variant of standard K-means technique that is Modified K-means technique. The Modified K-means technique has been developed to give better results as compared to existing K-means, Hierarchical and Cobweb techniques. This work also includes encryption and decryption of the formed clusters using AES algorithm to provide privacy to the data while transferring over networks. Experimental result proves that the performance of Modified K-means algorithm is better as compared to the existing K-Means and better than the hierarchical and Cobweb when tested on two datasets. K-Means and Hierarchical clustering is forming less number of clusters. In contrast, Cobweb is forming many clusters, which create memory issues. Therefore, Modified K-means forms an appropriate number of clusters in an organized manner and also takes minimum amount of time.