基于SSA优化神经网络的企业质量指标预测研究  

Research on Enterprise Quality Index Prediction Based on SSA Optimization Neural Network

作　　者：袁嬉莹 

机构地区：[1]上海理工大学管理学院,上海

出　　处：《建模与仿真》2023年第4期3909-3917,共9页Modeling and Simulation

摘　　要：由于目前生产过程中的监控体系不完整,生产方法落后,一直按照出现问题之后再解决问题的处理方案,且目前产品现代化生产过程中充满了不确定性和复杂性,因此这种生产管理方式不能够及时检测出生产过程中存在的质量问题。采用麻雀搜索算法对传统BP神经网络进行权值和阈值全局优化,提升对于卷烟制造中制丝环节的预测精准度,并将模型与传统BP神经网络通过损失函数进行对比,并引入KNN模型、SVM模型、Logistic回归分类模型和QDA二次判别分析模型进行对比,通过对上述五种模型的评估指数进行对比,证明了bp神经网络分类效果的优越性,而麻雀搜索算法优化BP神经网络在烟草生产过程中对质量管理预测具有高效性和准确性。Since the current monitoring system in the production process is incomplete and the production method is backward, it has been following the processing scheme of solving problems after they oc-cur, and the modern production process of current products is full of uncertainty and complexity, so this production management method is not able to detect the quality problems in the production process in time. The sparrow search algorithm is used to globally optimize the weights and thresh-olds of traditional BP neural network to improve the prediction accuracy for the filament making process in cigarette manufacturing, and the model is compared with traditional BP neural network by loss function, and KNN model, SVM model, logistic regression classification model and QDA quadratic discriminant analysis model are introduced to compare the above five models. The eval-uation indices were compared to prove the superiority of bp neural network classification effect, and the sparrow search algorithm optimized BP neural network has high efficiency and accuracy for quality management prediction in tobacco production process.

关 键 词：BP神经网络 LOGISTIC回归 生产管理方式 二次判别分析 损失函数 指标预测 全局优化 分类模型 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]