
科学技术与工程

中国技术经济学会
PemSet-FIM: Two-stage Adaptive Algorithm for Frequent Itemset Mining over Set-valued Data
Jing-zhe JIAO , Xiao-hui LI , Xing ZHANG , Jie LAN
Science Technology and Engineering ›› 2026, Vol. 26 ›› Issue (31) : 13563 -13575.
To address the problems of low-quality candidate sets, severe noise accumulation and insufficient frequency estimation accuracy of existing frequent itemset mining methods under local differential privacy for high-dimensional sparse data, a two-stage adaptive perturbation algorithm was proposed. With the domain-adaptive perturbation method, the high-precision mining mechanism of frequent itemsets in high-dimensional sparse scenarios was investigated. The experimental results show that dynamically selecting perturbation mechanisms in the candidate generation stage can effectively reduce the noise of initial candidates and improve the quality of prefix filtering; adaptively selecting mechanisms according to subdomain size in the frequency estimation stage can significantly suppress local variance and optimize global frequency reconstruction. Experiments on four real-world datasets demonstrate that the proposed algorithm outperforms existing methods in frequency estimation accuracy, frequent item discovery capability and operational efficiency. On highly sparse datasets, the Precision and Recall are each improved by up to 15%. It can be concluded that the proposed two-stage adaptive perturbation algorithm can significantly boost the frequent itemset mining performance on high-dimensional sparse datasets while guaranteeing local differential privacy.
local differential privacy / adaptive subdomain perturbation / frequent itemset mining / high-dimensional data / randomized response
/
| 〈 |
|
〉 |