Please wait a minute...
工程设计学报
工程设计学报  2022, Vol. 29 Issue (1): 20-27    DOI: 10.3785/j.issn.1006-754X.2022.00.016
保质设计     
改进MDSMOTE与PSO-SVM在汽车组合仪表分类预测中的应用
肖圳1, 何彦1, 李育锋1, 吴鹏程1, 刘德高2, 杜江2
1.重庆大学 机械传动国家重点实验室， 重庆 400030
2.重庆矢崎仪表有限公司， 重庆 401123
Application of improved MDSMOTE and PSO-SVM in classification prediction of automobile combination instrument
XIAO Zhen1, HE Yan1, LI Yu-feng1, WU Peng-cheng1, LIU De-gao2, DU Jiang2
1.State Key Laboratory of Mechanical Transmissions, Chongqing University, Chongqing 400030, China
2.Chongqing Yazaki Meter Co., Ltd., Chongqing 401123, China
 全文: PDF(2122 KB)   HTML
摘要： 汽车组合仪表生产过程中质检项目多且检测时间长，这在一定程度上制约了其生产效率的进一步提升。为此，提出一种基于改进最远点合成少数类过采样技术（max distance synthetic minority over-sampling technique，MDSMOTE）的支持向量机（support vector machine， SVM）分类预测方法。首先，结合专家经验对汽车组合仪表的原始生产数据进行特征筛选，并在MDSMOTE中引入类不平衡率IR，以对所筛选的特征数据进行扩充；然后，利用粒子群优化（particle swarm optimization， PSO）算法对SVM的误差惩罚因子C和核函数参数γ进行优化；最后，建立优化的SVM分类预测模型，并对汽车组合仪表进行分类。通过与其他分类预测模型在不同数据集上的预测结果进行对比可知，基于改进MDSMOTE的SVM分类预测模型的准确率、F值和几何平均值等评价指标均优于其他模型。所提出方法在汽车仪表产品分类上表现出较强的泛化能力和稳定性，可为仪表制造企业生产效率的提升提供有效参考。
Abstract: The numerous quality inspection items and long inspection time in the production process of automobile combination instruments have restricted the further improvement of production efficiency to a certain extent. To this end, a support vector machine (SVM) classification prediction method based on the improved max distance synthetic minority over-sampling technique (MDSMOTE) was proposed. Firstly, the feature selection for the original automobile combination instrument production data was carried out combined with the expert experience, and the class imbalance rate IR was introduced into the MDSMOTE to expand the selected feature data; then, the error penalty factor C and the kernel parameter γ of the SVM were optimized by the particle swarm optimization (PSO) algorithm; finally, an optimized SVM classification prediction model was established to make classifications for the automobile combination instruments. Compared with the prediction results of other classification prediction models on different data sets, the SVM classification prediction model based on the improved MDSMOTE was superior to other models in terms of the evaluation indexes as accuracy, F value and geometric mean value. The proposed method shows strong generalization ability and stability in the classification of automotive instrument products, which can provide an effective reference for the improvement of production efficiency of instrument manufacturers.
收稿日期: 2021-03-09 出版日期: 2022-02-28
CLC:  U 463.7  
基金资助: 重庆市技术创新与应用示范专项重点示范项目（cstc2018jszx-cyzdX0147）
通讯作者: 何 彦（1981—），女，重庆人，教授，博士生导师，博士，从事数字化制造与装备智能化、绿色设计与制造研究， E-mail：heyan@cqu.edu.cn，https：//orcid.org/0000-0002-6287-1130     E-mail: heyan@cqu.edu.cn
作者简介: 肖 圳（1995—），男，重庆人，硕士生，从事智能制造和产品质量预测研究，E-mail：xiaozhen@cqu.edu.cn，https：//orcid.org/0000-0003-1153-7754；
服务　　
把本文推荐给朋友
加入引用管理器
E-mail Alert
RSS
作者相关文章　　

引用本文:

肖圳, 何彦, 李育锋, 吴鹏程, 刘德高, 杜江. 改进MDSMOTE与PSO-SVM在汽车组合仪表分类预测中的应用[J]. 工程设计学报, 2022, 29(1): 20-27.

XIAO Zhen, HE Yan, LI Yu-feng, WU Peng-cheng, LIU De-gao, DU Jiang. Application of improved MDSMOTE and PSO-SVM in classification prediction of automobile combination instrument[J]. Chinese Journal of Engineering Design, 2022, 29(1): 20-27.

链接本文:

https://www.zjujournals.com/gcsjxb/CN/10.3785/j.issn.1006-754X.2022.00.016    或    https://www.zjujournals.com/gcsjxb/CN/Y2022/V29/I1/20

No related articles found!