In this chapter, a new algorithm for performing privacy preserving collaborative filtering is proposed by extending the conventional crisp k-member coclustering model into a fuzzy variant.Although the conventionalmethod anonymizes a co-occurrence data matrix by clustering objects into crisp object clusters in conjunction with fuzzy item membership estimation, the new algorithm constructs an anonymized data matrix considering fuzzy partition of objects. Because fuzzy partition is expected to be robust against outliers and extract homogeneous clusters, the proposed method can achieve k-anonymization with fewer information losses than the conventional crisp one. The applicability of the proposed algorithm to collaborative filtering task is demonstrated in a numerical experiment with a real-world purchase history data set.
CITATION STYLE
Honda, K., Kawano, A., & Notsu, A. (2015). A greedy fuzzy k-member co-clustering algorithm and collaborative filtering applicability. Smart Innovation, Systems and Technologies, 30, 39–50. https://doi.org/10.1007/978-3-319-13545-8_3
Mendeley helps you to discover research relevant for your work.