Global Optimization strategies for two-mode clustering

Two-mode clustering is a relatively new form of clustering that clusters both rows and columns of a data matrix. To do so, a criterion similar to k-means is optimized. However, it is still unclear which optimization method should be used to perform two-mode clustering, as various methods may lead to non-global optima. This paper reviews and compares several optimization methods for two-mode clustering. Several known algorithms are discussed and a new, fuzzy algorithm is introduced. The meta-heuristics Multistart, Simulated Annealing, and Tabu Search are used in combination with these algorithms. The new, fuzzy algorithm is based on the fuzzy c-means algorithm of Bezdek (1981) and the Fuzzy Steps approach to avoid local minima of Heiser and Groenen (1997) and Groenen and Jajuga (2001). The performance of all methods is compared in a large simulation study. It is found that using a Multistart meta-heuristic in combination with a two-mode k-means algorithm or the fuzzy algorithm often gives the best results. Finally, an empirical data set is used to give a practical example of two-mode clustering.