计算机工程

• 人工智能及识别技术 • 上一篇    下一篇

基于粒子群优化和判别熵信息的基因选择算法

关 健,韩 飞,杨善秀   

  1. (江苏大学计算机科学与通信工程学院,江苏 镇江 212013)
  • 收稿日期:2012-10-29 出版日期:2013-11-15 发布日期:2013-11-13
  • 作者简介:关 健(1987-),男,硕士研究生,主研方向:数据挖掘,智能计算;韩 飞,副教授、博士;杨善秀,硕士研究生
  • 基金项目:
    国家自然科学基金资助项目“编码先验约束的高维小样本数据处理方法的研究”(61271385)

Gene Selection Algorithm Based on Particle Swarm Optimization and J-divergence Entropy Information

(School of Computer Science and Telecommunication Engineering, Jiangsu University, Zhenjiang 212013, China)   

  1. (School of Computer Science and Telecommunication Engineering, Jiangsu University, Zhenjiang 212013, China)
  • Received:2012-10-29 Online:2013-11-15 Published:2013-11-13

摘要: 为了以较少冗余的特征基因得到较高的分类准确率,提出一种基因选择算法。通过分析基因对不同类别间的判别熵信息,剔除大量的冗余基因,以形成一个初选基因库。在初选基因库中,运用粒子群优化算法结合基因组,对不同类别间的判别熵信息和样本分类准确率进行最优基因子集选择。在2组基因微阵列数据上的实验结果表明,该算法不仅能够获取较少冗余的可解释基因子集,而且对最终选择出的特征基因也能获得较高的样本识别率。

关键词: 粒子群优化, 判别熵, 微阵列数据, 基因选择, 极端学习机, 先验信息

Abstract: In this paper, a new hybrid feature selection method is proposed to select informative genes with little redundancy and high classification accuracy. Most redundant genes are weeded out by J-divergence entropy between different categories to form the pool of genes, then within which Particle Swarm Optimization(PSO) algorithm is used as the optimal search algorithm to combine J-divergence entropy and classification rate together to select the optimal gene subset. The algorithm is tested on two common used microarray-data, which show the proposed method selects less redundant genes with more interpretability as well as increases prediction accuracy.

Key words: Particle Swarm Optimization(PSO), J-divergence entropy, microarray data, gene selection, extreme learning machine, prior information

中图分类号: