时间分层组合预测法的改进及其在进口贸易中的应用

针对现有时间分层组合预测法中协方差估计存在的不足,提出一种Shrinkage方法对协方差进行修正,以增强协方差估计的稳定性。Monte Carlo模拟结果表明:Shrinkage时间分层组合预测法的预测准确性较现有方法有所提高,且对季节因素不敏感;将该方法应用于中国进口贸易的预测中,结果显示新方法可显著提升各层预测效果,且对较高层序列的调整效果更佳;Shrinkage分层组合预测法能为海关等相关部门提供一种可针对不同时间维度进行同时性预测的新思路。

The Improvement of Time Hierarchy Combination Forecasting Method and Its Application in Import Trade

Aiming at shortcomings of covariance estimation of existing Time Hierarchy Combination Forecasting ,this paper proposes a Shrinkage method to revise sample covariance and increase the stability of covariance estimation .The prediction accuracy of Shrinkage Time Hierarchy Combination Forecasting method is higher than those of existing methods and is not sensitive to seasonal trends . Applying Shrinkage Time Hierarchy Combination Forecasting to China's import trade forecasting ,the result reveals that :T he new method significantly enhances the prediction accuracy at each levels and the accuracy improvement of higher‐level time series is better . Shrinkage Time Hierarchy Combination Forecasting provides a new thought for custom and relevant departments to implement simultaneous forecasting for different time dimensions .