Privacy preserving fuzzy co-clustering with distributed cooccurrence matrices
Daiji Tanaka, Toshiya Oda, Katsuhiro Honda, Akira Notsu · 2014
Privacy preserving data mining is a promising topic for utilizing various personal information without fear of information leaks. Fuzzy co-clustering is a fundamental technique for summarizing mutual cooccurrence information among objects and items, and has been demonstrated to be useful in such applications as document analysis and collaborative filtering. In this paper, a secure framework for privacy preserving fuzzy co-clustering is proposed for handling both vertically and horizontally distributed cooccurrence matrices. Personal observation stored in each site is summarized into co-cluster structures with an encryption operation. The advantage of utilizing distributed cooccurrence matrices is demonstrated in several numerical experiments.