Improving healthcare using privacy preserving association rule mining in distributed healthcare data

Abstract

The trend of data mining in healthcare has increased due to the digitalization of hospitals with electronic health records (EHR) system. The data stored at EHR systems are valuable assets for medical research. Association between disease and patient’s symptoms facilitate the doctors in taking healthcare decisions. The accuracy of decision can be enhanced by association rule mining on distributed healthcare data. The prerequisite is to share healthcare data by all collaborative EHR systems. Disclosing patient’s healthcare data for collaborative data mining may cause privacy issues. Privacy preserving distributed data mining solves this problem by achieving privacy and collaborative data mining results. Existing cryptography solutions provide privacy with higher computation and communication cost. In this paper, we propose an efficient approach for finding association rules in distributed horizontally partition healthcare data with comparatively efficient communication and computation cost. The theoretical and practical evaluation shows that our approach is efficient, scalable and outperforms existing approach. At last, we have shown the real application of proposed approach for breast cancer prediction based on symptoms of patients.