中文摘要：

从三维微阵列数据集挖掘出的三维聚类，能够分辨出与某些表现型（如疾病）相关的样本，而且能找出与这些表现型相关的候选基因．当挖掘3D微阵列数据矩阵中的3D聚类时往往要考虑同时优化几个目标，而且这些目标经常相互冲突，因此，应用多目标进化算法来求解GST数据集中的3D聚类是可行的．本文基于∈-支配和σ选择操作策略，提出一个新奇的多目标进化三维聚类算法来挖掘三维微阵列数据集中的3D聚类．通过在酵母细胞周期调控基因数据集上应用此算法，实验结果表明我们的方法能发现具有重大意义的高质量的3D聚类．

英文摘要：

3D （three-dimensional） clusters mining from gene-sample-time （simply GST） microarray data can identify the samples corresponding to some phenotypes, such as diseases, and find the candidate genes correlated to phenotypes. When mining 3D clusters in 3D microarray data matrix, several objectives have to be optimized simultaneously, and often these objective are in conflict with each other. Therefore, it is very available to use a multi-objective evolutionary algorithms （MOEA） for finding 3D clusters in GST data. Based on E-dominance and sigma select strategy, this paper proposes a novel multi-objective evolutionary 3D clustering algorithm to mine 313 cluster from 3D microarray data. Experimental results on yeast cell cycle dataset show that our approach can find significant 3D clusters of high quality.