An Iterative Association Rule Mining Framework to K-Anonymize a Dataset

Authors

Michael Hayes, Western UniversityFollow
Miriam A M Capretz, Western UniversityFollow
Jefferey Reed, Western UniversityFollow
Cheryl Forchuk, Western UniversityFollow

Document Type

Article

Publication Date

2012

Volume

1

Issue

4

Journal

ASE Science Journal

First Page

179

Last Page

194

Abstract

Preserving and maintaining client privacy and anonymity is of utmost importance in any domain and specially so in healthcare, as loss of either of these can result in legal and ethical implications. Further, it is sometimes important to extract meaningful and useful information from existing data for research or management purposes. In this case it is necessary for the organization who manages the dataset to be certain that no attributes can identify individuals or group of individuals. This paper proposes an extendable and generalized framework to anonymize a dataset using an iterative association rule mining approach. The proposed framework also makes use of optional domain rules and filter rules to help customize the filtering process. The outcome of the proposed framework is a preprocessed dataset which can be used in further research with confidence that anonymity of individuals is conserved. Evaluation of this research will also be described in the form of a case study using a test dataset provided by the Lawson Health Research Institute in London, Ontario, Canada as a part of their Mental Health Engagement Network (MHEN) study.