面向高维不平衡医学数据的特征选择算法  

Feature Selection Algorithm for High-dimensional Unbalanced Medical Data

作　　者：苏璇 王远军[1] SU Xuan;WANG Yuanjun(Institute of Medical Imaging Engineering,University of Shanghai for Science and Technology,Shanghai 200093,China)

机构地区：[1]上海理工大学医学影像工程研究所,上海200093

出　　处：《小型微型计算机系统》2024年第2期309-318,共10页Journal of Chinese Computer Systems

基　　金：上海市自然科学基金项目(18ZR1426900)资助。

摘　　要：基于传统机器学习分类算法对影像组学的高维不平衡数据分类结果不理想的问题,本文提出一种改进海洋捕食者的不平衡特征选择算法.首先,对海洋捕食者算法MPA算法进行改进,引入精英反向矩阵增加算法迭代后期的种群多样性,引入新的CF参数改善算法的收敛速度与精度,同时合理分配原始参数分布和取值来满足算法在不同阶段的搜索需求;接着针对不平衡数据引入新的目标函数来帮助MPA算法收敛到更优的特征子集.最后,基于G-means的精英反向海洋捕食者算法GEMPA算法在14个基础测试函数上进行测试并在12个公开数据集上与MPA,基于K个最近邻相关性的在线特征选择算法K-OFSD以及其余的6种元启发式算法GA、PSO、CSO、SSA、SCA和MFO对比分析.以平均F-measure值,平均特征数量,平均运行时间为评估指标,通过实验可知GEMPA算法能够快速搜索到分类精度最高的特征子集,降低高维数据的冗余度,针对改善高维不平衡数据分类问题有很好的发展潜力.Based on the problem of unsatisfactory results of traditional machine learning classification algorithms for high-dimensional imbalanced data classification in radiomics,an imbalanced feature selection algorithm for marine predators MPA was proposed.First,the MPA algorithm was improved.The elite opposition matrix was introduced to increase the population diversity in the later iteration of the algorithm.The new CF parameter was introduced to improve the convergence speed and accuracy of the algorithm,and the original parameter distribution and value were reasonably allocated to meet the needs of the algorithm in different stages.Then,a new objective function was introduced for unbalanced data to help the MPA algorithm converge to a better feature subset.Finally,the elite opposition solution MPA algorithm based on G-means GEMPA was tested on 14 basic test functions and compared with MPA,online feature selection based on the dependency in K nearest neighbors K-OFSD,and the other six meta-heuristic algorithms on 12 public datasets,such as GA、PSO、CSO、SSA、SCA and MFO.Taking the average F-measure value,the average number of features,and the average running time as the evaluation indicators,it can be seen through experiments that the GEMPA algorithm can quickly search for the feature subset with the highest classification accuracy,reduce the redundancy of high-dimensional data.GEMPA has good development potential in improving high-dimensional imbalanced data classification issue.

关 键 词：特征选择 高维不平衡 海洋捕食者算法 反向学习 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] TP181[自动化与计算机技术—计算机科学与技术]