中文说明：应用背景 虽然传统的基因选择方法已经能够取得很好的效果，选出的基因子集有利于后续样本分类，但是这些方法主要考虑数据方差和分布的相关性，从而选出的基因可解释性较差且冗余度较高。为了获得最小冗余可解释的基因子集，本文在充分考虑基因类别灵敏度 (Gene to class sensitivity,GCS) 信息和基因调控 (Gene regulation,GR) 信息的基础上，利用二进制微粒群算法(BPSO)和极限学习机(ELM)进行基因选择。 关键技术该方法提出用PSO优化K均值聚类，并在此基础上确定初始化种群，随后运用两个BPSO分别结合GCS和GR信息进行进行选择。在多个基因表达谱数据上的实验结果表明，相比BPSO-GCS-ELM和其他一些经典的基因选择方法，该方法能够选出更紧凑且分类能力更强的基因子集。
English Description：
Application backgroundAlthough the traditional method of gene selection has been able to achieve good results, the selected gene subset is conducive to the follow-up sample classification, but these methods mainly consider the correlation between the data and the distribution of variance, so that the selected genes can be interpreted as poor and redundant. In order to obtain the minimal redundancy of the gene subset, the sensitivity of (Gene , to ,, GCS, class , sensitivity), ELM (Gene ) (BPSO) (regulation) (GR) was used for gene selection.Key TechnologyThe method proposed in this paper is to optimize the K mean clustering by PSO, and then determine the initial population, and then select the GCS and GR based on the two BPSO. Experimental results on multiple gene expression profiles show that the proposed method is capable of selecting a more compact and classification ability for a more compact subset o