Privacy preserving data mining using general regression auto-associative neural network: Application to regression problems

Abstract

Data mining has proved its significance in various areas like healthcare, counter-terrorism etc. But it has several times raised issues concerning privacy, legality and ethics. Thus, a need for privacy preserving data mining arose. While preserving privacy one thing that should be made sure is that the accuracy of the final predictions should not suffer drastically. This paper proposes a novel General Regression Auto- Associative Neural Network (GRAANN) for privacy preservation. Then, General Regression Neural Network (GRNN) is applied for data mining purpose, leading to GRAANN + GRNN hybrid. The hybrid is tested on five benchmark datasets. From the accuracy of the predictions made it can be concluded that GRAANN can be used as optimum technique for privacy preservation.