A Density Peaking Clustering Algorithm for Differential Privacy Preservation

The privacy protection problem in data mining has received increasingly attention and is a hot topic of current research. To address the problems large accuracy loss instability clustering results algorithms under differential requirements, density peak algorithm for (DP-chDPC) proposed. Firstly, original DPC improved, by using dichotomy method to automatically determine truncation distance avoid subjectivity manual selection, setting threshold local center offset obtain center, which overcomes uncertainty select based on decision graph. Then, noise added Laplace mechanism realize during analysis. Finally, Chebyshev used replace Euclidean calculate matrix, reduces interference after adds noise, accuracy, so that stability improved. experimental show DP-chDPC can effectively reduce are more stable.