Abstract: This paper introduces the idea “first test and then generation” to address the efficiency issue of clustering 3D gene expression data. Based on the proposed concept of closed similarity， TESTER， a novel efficient algorithm， is designed. Meanwhile， several effective pruning rules are devised to avoid the time-consuming global closure test. Both theoretic analysis and experimental results prove that TESTER outperforms RSM and CubeMiner， the two state-of-the-art homogeneous algorithms.

Key words: 3D gene expression data, closed similarity, clustering

摘要： 研究了三维基因表达数据聚类的效率问题，在三维基因表达数据聚类过程中引入了“先验证后生成”的思想。基于提出的封闭相似性概念，设计了一种新的高效算法TESTER，采用多个有效的削减规则避免代价很高的全局封闭性检验，提高了效率。理论分析和实验结果表明，TESTER算法的性能优于目前最好的同类算法RSM和CubeMiner。

关键词: 三维基因表达数据, 封闭相似性, 聚类