Privacy and Clustering Based Online Feature Selection with Public Auditing

Feature selection is essential topic in data mining. Although its importance, most studies of feature selection are limited to batch learning. The online feature selection is used to make accurate prediction for using small number and fixed number of active features . We deal with this challenge by studying scarcity regularization and truncation techniques. we estimate the performance of proposed algorithms for online feature selection and its applications .we propose Two-Gaussian algorithm for clustering a search result. Our predicted informations are separately grouped in the basis of classification. This clustering technique done by using Two Gaussian mixtures algorithm. And we implement Blowfish algorithm for reduce privacy issues. All informations are stored in our system as an encrypted format. And also we implement public auditing for audit a user contents. Because it is used to avoid fake informations.