摘要
介绍了组合预测的方法,并利用最优组合和递归方差倒数方法对组合预测方法进行改进;提出通过GMDH方法首先对影响经济预测模型的各变量进行筛选然后再建立回归模型、神经网络模型等单项预测模型的思想;最后结合GMDH方法建立的时间序列模型,建立正权重组合预测模型.
The combined forecasting and the way of improving such forecasting methodology through optimized combination and recursion were introduced. The GMDH was proposed to be applied to select variables relevant to economic forecasting models before establishing a single forecasting model such as regression model and neural network model. And the final step was to establish a combined model for forecasting based on positive weight by using the time series model, which was established according to the GMDH.
出处
《经济数学》
北大核心
2010年第1期85-91,共7页
Journal of Quantitative Economics
关键词
正权重
组合预测
GMDH
神经网络
positive weight
combined forecasting
GMDH
neural network
