一种微阵列数据降维新方法

Novel method for microarray data dimension reduction

提出一种二阶段并行基因选择方法(TPM),可以获得最优特征子集。针对以往算法易于陷入局部极值的不足,提出了一种模糊多种群粒子群(FMP),可以有效地扩展搜索空间。通过在leukemia、colon、breast cancer、lung carcinoma、brain cancer五个数据集上的测试,验证了本文方法不仅可以获得更优特征子集和更高的分类精度,而且可以选择尺寸更小的特征子集。本文的研究成果可为基因表达领域提供一种新的思路。

A two stage parallel gene selection method （TPM） for obtaining the optimal feature subset is proposed. A fuzzy multi-swarm particle optimization （FMP） is also proposed to extend the searching spaces, to overcome the problem of traditional algorithm to be locked to local optimum. The performance of the TMP is evaluated on five microarray datasets （leukemia dataset, colon dataset, breast cancer dataset, lung eareinoma dataset and brain cancer dataset）. The comparison results show that the proposed method not only gets better quality of feature subset and higher classification accuracy, but also generates smaller feature subsets. The results of this study could provide a new idea to the field of gene expression.