摘要
提出基于多目标鲸鱼优化(multi-objective whale optimization algorithm, MWOA)的产品关键质量特性(特征)识别方法。首先,针对非平衡制造过程数据,将识别模型构建为最大化G-mean指标和最小化特征数的多目标特征选择问题。其次,提出群智能优化算法MWOA求解模型。MWOA针对特征选择提出一种新的多样性帕累托排序策略,能够对解进行优劣排序的同时保证群体多样性。同时,在MWOA中嵌入变异操作以解决鲸鱼优化易陷入局部最优的缺点。选取3组非平衡制造过程数据验证算法有效性。实验结果表明,所提算法在数据非平衡条件下能够有效识别关键质量特性。
In this paper,a feature selection method based on a multi-objective whale optimization algorithm(MWOA)is proposed for the key quality characteristics(KQC)identification problem.First,the identification model is established as maximizing the G-mean metric and minimizing feature(quality)subset size for imbalanced manufacturing data.Second,a swarm-based optimization algorithm,i.e.,MWOA,is proposed to solve this model.MWOA adopts a new diversity Pareto sorting strategy,which sorts solutions in the swarm as well as keeps the swarm diversity for feature selection problems.Moreover,a mutation operator is embedded in MWOA as the search strategy of whale optimization can easily get trapped in the local optimum.The experimental results on 3 imbalanced manufacturing datasets illustrate that the proposed method can effectively identify KQCs.
作者
李岸达
何桢
王庆
LI An-da;HE Zhen;WANG Qing(School of Management,Tianjin University of Commerce,Tianjin 300134,China;College of Management and Economics,Tianjin University,Tianjin 300072,China)
出处
《系统工程》
CSSCI
北大核心
2019年第1期134-142,共9页
Systems Engineering
基金
教育部人文社会科学研究一般项目(19YJC630071)
国家自然科学基金(71661147003
71532008)
天津市教委科研计划项目成果(161072
基于非平衡数据的复杂产品关键质量特性识别研究)
关键词
关键质量特性识别
特征选择
多目标优化
鲸鱼优化算法
非平衡数据
key Quality Characteristics Identification
Feature Selection
Multi-Objective Optimization
Whale Optimization Algorithm
Imbalanced Data
