Privacy Preserving Market Basket Data Analysis

Randomized Response techniques have been empirically investigated in privacy preserving association rule mining. However, previous research on privacy preserving market basket data analysis was solely focused on support/ confidence framework. Since there are inherent problems with the concept of finding rules based on their support and confidence measures, many other measures (e.g., correlation, lift, etc.) for the general market basket data analysis have been studied. How those measures are affected due to distortion is not clear in the privacy preserving analysis scenario. In this paper, we investigate the accuracy (in terms of bias and variance of estimates) of estimates of various rules derived from the randomized market basket data and present a general framework which can conduct theoretical analysis on how the randomization process affects the accuracy of various measures adopted in market basket data analysis. We also show several measures (e.g., correlation) have monotonic ...

• Type: Conference
• Year: 2007
• Where: PKDD
• Authors: Ling Guo, Songtao Guo, Xintao Wu
• Keywords: Basket Data Analysis, Market Basket Data, PKDD 2007, Randomized