T Evaluating the Classification Accuracy of Data Mining Algorithms for Anonymized Data

Recent advances in hardware technology have increased storage and recording capability with regard to personal data on individuals. This has created fears that such data could be misused. To alleviate such concerns, data was anonymized and many techniques were recently proposed on performing data mining tasks in ways which ensured privacy. Anonymization techniques were drawn from a variety of related topics like data mining, cryptography and information hiding. Data is anoymized through methods like randomization, k-anonymous, l-diversity. Several privacy preserving data mining algorithms are available in literature. This paper investigates the classification accuracy of the data with and without k-anonymization to compare the efficiency of privacy preserving mining. The classification accuracy is evaluated using k nearest neighbor, J48 and Bagging.