An efficient numerical method for the prediction of clusters using K-means clustering algorithm with bisection method

Abstract

The development of modern IT-based analysis methods, data mining, has been outstanding over the last decade. Using computers to analyze masses of information to discover trends and patterns. The current trend in business collaboration shares the data and mines results to gain mutual benefit. The main goal of the work is to introduce a bisection method which is capable of transforming a non-anonymous data set into adult data set. In this model, transform a table so that no one can make high probability association between records in the table and the corresponding entities. In order to achieve these goals we are implemented a bracket rule identifier for the prediction of the cluster. For this a suitable metric has been developed to estimate information loss by suppression which works well for both numeric and categorical data. © 2012 Springer-Verlag.