基于磷虾-黑洞算法的医学数据特征选择研究  

Feature Selection of Biomedical Data Based on Krill Herd and Black Hole Algorithm

作　　者：张戈[1] 彭振 侯金翠 阎朝坤[1] ZHANG Ge;PENG Zhen;HOU Jincui;YAN Chaokun(School of Computer and Information Engineering,Henan University,Henan Kaifeng 475004,China;CMCC-Online Company,Zhengzhou 450007,China)

机构地区：[1]河南大学计算机与信息工程学院,河南开封475004 [2]中移在线服务有限公司,郑州450007

出　　处：《河南大学学报（自然科学版）》2022年第6期690-698,共9页Journal of Henan University:Natural Science

基　　金：国家自然科学基金资助项目(61802114,61802113,61972134);中国博士后科学基金(2020M672212);河南省重点研发与推广专项科技攻关项目(202102210173,212102210091)。

摘　　要：从高维的生物医学数据中探索发现与疾病相关的基因是目前的热点研究问题,但是大部分生物医学数据具有许多与寻找疾病基因不相关或冗余特征,很难直接投入使用.针对这个问题,提出了一种自适应双种群混合磷虾黑洞算法(modified binary krill herd and black hole algorithm, MBKHA).该算法将改进的二进制磷虾算法与二进制黑洞算法相结合,磷虾算法负责寻找更优的解集,黑洞算法负责加快算法收敛,通过使用自适应划分规则动态调控种群中磷虾个体和恒星个体的数量,从而实现两个算法优势互补.基于5个公开医学微阵列数据集,从多个指标比较了提出的方法和其他特征选择算法的性能,实验结果表明该方法在特征选择上具有更好的性能.The discovery of disease-related genes from high-dimensional biomedical data is a hot research issue, but most biomedical data have many irrelevant or redundant features, difficult to use directly. An adaptive dual-population hybrid krill black hole algorithm(modified binary krill herd and black hole algorithm, MBKHA) was proposed to solve this problem. The algorithm combined the improved binary krill algorithm with the binary black hole algorithm. The krill algorithm was responsible for improving the solution’s quality, and the black hole algorithm was responsible for accelerating the convergence speed. The number of different individuals in a population were controlled using adaptive partitioning rules, so the krill algorithm and the black hole algorithm are dynamically adjusted. Based on five public medical microarray data sets, the proposed method’s performance and other feature selection algorithms are compared from multiple indicators. Experimental results show that this method has better performance in feature selection.

关 键 词：特征选择 改进的二进制磷虾算法 二进制黑洞算法 生物医学数据 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]