The results presented there, regarding an updated version of the paper's results, were much better than the six algorithms chosen as part of the experiment regarding volume and coverage while maintaining the residue value bellow the estipulated delta.
Also the time taken to find such high quality solutions were just slightly worse than the Cheng and Church algorithm, that is a constructive heuristic, and better than the other approaches. This position the SwarmBcluster as a nice choice to find high quality biclusters on any kind of set of data.
Although nowadays the biclustering algorithms are mainly used as a tool to find out previously unknown relations on microarray datasets, in my point of view this type of clustering will represent a major role in the future of data mining.
On the next few days I'm going to ellaborate a post about some topics in which the bicluster may be useful or interesting.
Nenhum comentário:
Postar um comentário