慢性肾病预测的多目标特征选择  

Multi-objective feature selection for chronic kidney disease prediction

作　　者：陈陆爽 周晖[1] CHEN Lu-shuang;ZHOU Hui(School of Information Science and Technology,Nantong University,Nantong 226019,China)

机构地区：[1]南通大学信息科学技术学院,江苏南通226019

出　　处：《计算机工程与设计》2022年第3期698-705,共8页Computer Engineering and Design

基　　金：国家自然科学基金项目(61501264)。

摘　　要：针对慢性肾病(chronic kidney disease,CKD)致死率高、早期症状不明显的特征,结合互信息和皮尔逊相关系数两种评价准则提出一种慢性肾病预测的多目标特征选择模型。针对慢性肾病预测,提出多目标群集智能特征选择算法MCFS,所提算法在GWO的基础上采用精英反向学习、非线性控制参数和联想记忆策略3个改进算子。仿真结果表明,所提算法对CKD的预测准确率高,筛选出与CKD紧密相关的特征子集能力强,明显优于现有的CKD预测方法和其它特征选择算法,能够为CKD早期患者提供准确可靠的辅助诊断。In view of the characteristics of high mortality rate and unobvious early symptoms of chronic kidney disease,a multi-objective feature selection model based on mutual information and Pearson correlation coefficient for the prediction of chronic kidney disease was proposed.Aiming at the prediction of chronic kidney disease,a multi-objective swarm intelligent feature selection algorithm MCFS was proposed.Three improved operators including elite reverse learning,a nonlinear control parameter and associative memory strategy based on GWO were used.The simulation results show that the proposed algorithm has high prediction accuracy for CKD and has a strong ability to select feature subsets closely related to CKD,which is obviously superior to the existing CKD prediction methods and other feature selection algorithms,and it can provide accurate and reliable auxiliary diagnosis for early CKD patients.

关 键 词：慢性肾病 早期预测 多目标特征选择 帕累托沿 新型群智能算法 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]