我国月度进口量预测模型研究

用季节性单位根理论研究了我国月度进口量序列，通过模型形式的选择来解决单变量序列趋势规律的表达和季节性规律的识别问题。研究结果表明，在预测模型中不仅使用季节差分算子不合理，而且带有一阶差分和确定性季节变量的季节模型也会引起过差分，Ｂｏｘ－Ｊｅｎｋｉｎｓ季节模型效果较差，最佳模型是带有确定型趋势和季节哑变量的确定型模型（ＤＴＳＤ）。文中从多个角度对３类模型进行了严格检验，结果证实ＤＴＳＤ模型具有理论与实用价值。

Study on the Forecasting Models of Monthly Import in China

The monthly import series was studied using the theory of seasonal unit roots,and the problems of the expression of unvariate seasonal time series and its seasonalcharacteristics can be solved by selecting the model structures. The results showed thatnot only using the seasonal differences is unsuitable, but the models with the firstdifference and the seasonal dummy variables can resulting in over-differencing. The commonused Box-Jenkings seasonal models has the bad effects, the best models are the determinatemodels with the determinate time trend and the seasonal dummy variables (DTSD). Manystrict tests were made on the three types of the models above, the results showed thatDTSD models has the theoretical and practical values.