6
30

Utility-efficient Differentially Private K-means Clustering based on Cluster Merging

Abstract

Differential privacy is widely used in data analysis. State-of-the-art kk-means clustering algorithms with differential privacy typically add an equal amount of noise to centroids for each iterative computation. In this paper, we propose a novel differentially private kk-means clustering algorithm, DP-KCCM, that significantly improves the utility of clustering by adding adaptive noise and merging clusters. Specifically, to obtain kk clusters with differential privacy, the algorithm first generates n×kn \times k initial centroids, adds adaptive noise for each iteration to get n×kn \times k clusters, and finally merges these clusters into kk ones. We theoretically prove the differential privacy of the proposed algorithm. Surprisingly, extensive experimental results show that: 1) cluster merging with equal amounts of noise improves the utility somewhat; 2) although adding adaptive noise only does not improve the utility, combining both cluster merging and adaptive noise further improves the utility significantly.

View on arXiv
Comments on this paper