A Comparative Study on Three-mode Fuzzy Co-clustering Based on Co-occurrence Aggregation Criteria

Three-mode fuzzy co-clustering is a promising technique for analyzing relational co-occurrence information among three mode elements. This paper proposes a modified version of the three-mode fuzzy clustering for categorical multivariate data (3FCCM) algorithm, which was constructed with the aggregation criterion of three elements based on the fuzzy c-means (FCM) concept and often suffers from careful tuning of three independent fuzzification parameters. In the modified algorithm, a novel clustering criterion is proposed based on a probabilistic concept, where we can easily tune the fuzziness degree of three-mode fuzzy partition by comparing with the probabilistic standard. The characteristics of the two algorithms are discussed through comparative experiments such that the modified version is more useful in tuning the influences of the fuzziness parameters and is more promising in real applications than the conventional one.