An Improved Weighted-Feature Clustering Algorithm for K-anonymity

Abstract

Chiu proposed a clustering algorithm adjusting the numeric feature weights automatically for k-anonymity implementation and this approach gave a better clustering quality over the traditional generalization and suppression methods. In this paper, we propose an improved weighted-feature clustering algorithm which takes the weight of categorical attributes and the thesis of optimal k-partition into consideration. To show the effectiveness of our method, we do some information loss experiments to compare it with greedy k-member clustering algorithm.