期刊文献+

变权重组合预测模型的局部加权最小二乘解法 被引量:11

Local Weighting Least-square Method of Weight-varying Combination of Forecast Models
下载PDF
导出
摘要 随着科学技术的不断进步,预测方法也得到了很大的发展,常见的预测方法就有数十种之多。而组合预测是将不同的预测方法组合起来,综合利用各个方法所提供的信息,其效果往往优于单一的预测方法,故得到了广泛的应用。而基于变系数模型的思想研究了组合预测模型,将变权重的求取转化为变系数模型中系数函数的估计问题,从而可以基于局部加权最小二乘方法求解,利用交叉证实法选取光滑参数。其结果表明所提方法预测精度很高,效果优于其他方法。 With the development of technology, much improvement has been made in forecasting procedures and there are about tens of general approaches. In general, we can apply some different approaches for one forecasting problem, but we can not select the best one. Combination forecast models draw the information from some forecasting procedures by combining these methods as one model and have advantage on single forecasting model. Therefore, the combination forecast models have been widely studied. Based on the varying coefficient models, the paper studies the weight - varying combination of forecast models. Changeable weights can be solved by locally weighted least squares method and smoothing parameter was selected by Cross -Validation method. Finally, a practical example was clone and the results show that the effects of our approach are remarkable and the accuracy of forecasting is higher than other methods.
作者 李静
出处 《统计与信息论坛》 2007年第3期44-47,共4页 Journal of Statistics and Information
关键词 变权组合预测模型 变系数回归模型 局部加权最小二乘 交叉证实法 weight - varying combination of forecast models varying coefficient regression model local weighting least- square method cross- validation
  • 相关文献

参考文献14

二级参考文献43

共引文献268

同被引文献103

引证文献11

二级引证文献57

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部