摘要
针对单个神经网络模型易出现过拟合而导致泛化能力较弱的缺点,引入了神经网络集成方法,对传统的Bagging方法进行改进,提出了一种基于0.632误差聚类的Bagging方法。通过实验对比和假设检验,证实了该方法的优越性,并探讨了最佳聚类数目。最后,通过应用实例展示了利用集成神经网络进行产品完工期预测的全过程。实验结果显示,该方法明显地提高了预测精度。
To overcome the poor generalization ability of a single neural network resulting from the over fitting,neural network ensemble was introduced,and a new Bagging approach based on the cluster analysis of the 0.632 prediction error was proposed.Advantages of this proposed method was validated by experimental comparison and hypothesis test.The best number of clusters was also discussed.Finally,a case study was given to illustrate the whole steps to predict the product due date by using neural network ensemble.Results showed that this model could remarkably improve precision of prediction results.
出处
《计算机集成制造系统》
EI
CSCD
北大核心
2007年第11期2140-2144,共5页
Computer Integrated Manufacturing Systems
基金
国家973计划资助项目(2005CB724100)
国家自然科学基金资助项目(50675082)。~~
关键词
产品完工期
预测建模
神经网络集成
聚类分析
product due date
predictive modeling
neural network ensemble
cluster analysis
