基于BP神经网络的Stacking模型融合的光谱分类算法  

Spectral Classification Algorithm of Stacking Model Fusion Based on BP Neural Network

作　　者：林苗芳 LIN Miaofang(South China Normal University,Guangzhou 510631,China)

机构地区：[1]华南师范大学,广东广州510631

出　　处：《现代信息科技》2022年第4期91-94,99,共5页Modern Information Technology

摘　　要：文章介绍了Stacking集成学习方法、BP神经网络模型,分别讨论了单分类器的精度,特征提取与单分类器结合的算法的精度、Stacking集成学习算法的精度。实验测试表明,以决策树、LDA、朴素贝叶斯、随机森林为基学习器,以BP神经网络为元学习器所建立的Stacking算法是最佳的光谱分类算法,该算法稳定性强,学习效果好,平均准确率高达94%。This paper introduces Stacking ensemble learning method and BP neural network model,discusses respectively the accuracy of the single classifier,the accuracy of the algorithm combining feature extraction and single classifier,and the accuracy of the Stacking ensemble learning algorithm.The experimental test shows that the Stacking algorithm established by a base learning machine of the decision tree,LDA,Naive Bayes,random forest,and a meta learning machine of BP neural network,is the best spectral classification algorithm.The algorithm has strong stability,good learning effect,and the average accuracy rate is up to 94%.

关 键 词：Stacking模型融合 BP神经网络模型 多分类问题 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]