Mining Triclusters of Similar Values in Triadic Real-Valued Contexts
Analysis of polyadic data (for example, multi-way tensors and n-ary relations) becomes more and more popular task nowadays. While several datamining techniques exist for (numeric) dyadic contexts, their extensions to the triadic case are not obvious, if possible at all. In this work, we study development of the ideas of Formal Concept Analysis for processing three-dimensional data, namely the so called OAC-triclustering (from Object, Attribute, Condition). Among several known methods, we have reasonably selected the most effective one and used it to propose an algorithm NOAC-triclustering for mining triclusters of similar values in real-valued triadic contexts. We have also proposed a second simple algorithm, Tri-K-Means, based on clustering algorithm K-Means, for the purpose of comparison. The experimental part demonstrates application of the algorithms to both computer-generated and real-world data.