Influence of transitive closure complexity in FJP-based clustering algorithms

The difference of Fuzzy Joint Points (FJP) algorithm from other neighborhood-based clustering algorithms is that it uses the concept of fuzzy neighborhood when computing the neighborhood relations. Among the proposed methods, none of them is perfect for all aspects of clustering requirements. Although FJP algorithm has superiority by its advantages of robustness and optimal determination of the best cluster number, slow running time is its disadvantage. The main reason is that transitive closure matrix, computed on neighborhood relations, needs many operations. To reduce this defect, a novel bifurcated product method which needs approximately 2 / log n n times fewer operations than the method that the FJP uses is proposed in this study. The proposed method is investigated theoretically and experimentally for various data sets. The comparison with the FJP method for its known variant is performed.