一种基于粗糙k均值的双聚类算法

双聚类算法是为了发现基因表达数据矩阵中局部相似性而提出的新聚类方法。本文根据Cheng和Church的打分理论采用自底向上的策略，首先用粗糙k均值算法生成初始的基因数据块，再对这些数据块添加行和列，生成初始的双聚类。然后，删除初始的双聚类中一致性波动不好的行和列，从而得到最终的双聚类。实验表明，该算法能够高效地生成具有共表达水平的双聚类，更能找到一致波动水平很高的双聚类。

A Biclustering Algorithm Based on Rough K-means

Biclustering is a new branch of clustering which is proposed to discover the local homogenous pattern. Based on the score scheme which is proposed by Cheng and Church, this paper uses bottom-up strategy, firstly finds the gene data submatrix using Rough K-means algorithm, then adds some conditions and genes to the block, finally deletes some conditions and genes from the block if necessary, the final results are the biclusters. The experiment shows that this algorithm can efficiently find the co-regulation patterns from the exist gene expression data, especially those highly homogenous patterns.