Data Mining: Min–Max Normalization Based Data Perturbation Technique for Privacy Preservation

Abstract

Data mining system deals with huge volume of information which may include personal and sensitive information about the individuals such as bank credential details, financial records, health-related information, etc. Data mining process utilizes this information for analyzing purpose but privacy preservation of such sensitive data is very much crucial in data mining process in ordered to prevent the privacy about the individuals. In recent years, privacy preservation is an ongoing research topic because of the high availability of personal data which consist of private and sensitive information about the individuals. Data perturbation technique is a well-known data modification technique to preserve the privacy of sensitive values and achieves accurate data mining results. In data perturbation method, original data is perturbed (modified) before the data mining process begins. In the existing method, data modification is takes place by adding noise (Gaussian) to the original data. In this method, loss of data loss is little high, and to overcome such issue, proposed method is established. In this paper, min–max normalization-based data transformation method is used to protect the sensitive information in a dataset as well as to achieve good data mining results. The proposed method is applied on the adult dataset and the accuracy of the results is compared with Naive Bayes classification algorithm and J48 decision tree algorithm with minimum information loss by having high data utilization. The performance of the proposed method is examined with two major considerations like maintaining the accuracy of the data mining application along with privacy preservation of original data.