A novel algorithm of biclustering based on the association rules

Because of the ability of simultaneously capturing correlations among subsets of attributes (columns) and records (rows), biclustering is widely used in data mining applications such as biological data analysis, financial forecasting, and customer segmentation, etc. Since biclustering is known to be an NP-hard problem, biclusters are identified through heuristic approaches in most algorithms whose results are non-deterministic. A new algorithm based on association rules is proposed in this paper. It is deterministic and enables exhaustive discovery of coherent evolution biclusters. Furthermore, we propose the improved algorithm to avoid finding repetitive biclusters and this reduces the searching time. Finally, the improved algorithm is parallelized to accelerate the mining process, and significant speed-up ratio is achieved.