Abstract: Because microarray dataset contains a large number of columns and a small number of rows, mining the complete set of frequent closed patterns in microarray poses a great challenge for traditional frequent closed patterns mining algorithms which are based on column-enumerate. CARPENTER, based on row-enumerate, has addressed this problem. However, CARPENTER is inefficient by using the transposed table (TT) to mine the complete set of frequent closed patterns in microarray. A new algorithm, MFCPLG, is proposed inspired by CARPENTER, and a LG-tree structure is suggested. Several experiments are performed on real-life microarray data to show that the MFCPLG algorithm is faster than the CARPENTER algorithm.

Key words: microarray dataset, frequent closed patterns, MFCPLG, LG-tree

摘要： 微阵列数据集行少列多的特征，使得传统基于列枚举空间的算法应用于其中进行频繁闭合模式挖掘时其复杂性迅速增长。基于行枚举的CARPENTER算法较好解决了该问题。但CARPENTER算法使用映射转置表(TT)来完成频繁闭合模式完全集的挖掘效率不高。该文在CARPENTER算法基础上，提出LG-tree数据结构，并基于此结构提出挖掘频繁闭合模式的新算法MFCPLG。真实数据集的实验表明，MFCPLG算法的时间性能优于CARPENTER算法。

关键词: 微阵列数据集, 频繁闭合模式, MFCPLG, LG-tree

CLC Number: 

• TP391