An Improved Differential Privacy Algorithm Using Frequent Pattern Mining

Facing with a large number of long transactions, computation time-consuming of differentially private top-k pattern mining (DP-topkP) is particularly prominent especially when the minimum threshold min_sup gradually decreases or the transactions gradually increase. In this paper, we propose an improved DP-topkP algorithm. Inspired from truncated transactions, a mechanism of length selection is firstly exploited in order to predispose the transaction database. Then, the closed frequent pattern is adopted to reduce the size of the candidate frequent item set obtained by the FP-Growth algorithm. Extensive experiments on both pumsb-star and kosarak datasets demonstrate the effectiveness and feasibility of the improved DP-topkP compared with DP-topkP.