| 摘要: |
| 为提高农产品市场价格的预见性,及早采取措施减缓价格波动,以全国西红柿月度批发市场价格为预测目标,综合利用季节虚拟变量法、Census X12法、移动平均比率法、Holt-Winters季节指数平滑法、SARIMA法等建立短期预测模型,并根据模型预测误差大小赋予不同的权重值,从而建立组合预测方法。实证分析结果表明:单一模型预测误差波动较大,总体上随着预测周期变长精度下降。在2009年的评估预测中,所建立的5个单一短期预测模型平均绝对误差百分比(MAPE)为10%左右,其中Holt-Winters季节指数平滑法建立的短期预测模型精度最高,MAPE为6.81%。如果预测提前期为3个月,SARIMA模型的预测精度更高,准确率达到95%以上。在实证分析的基础上,采用组合预测方法对2010年西红柿价格进行了预测。 |
| 关键词: 农产品 市场价格 预测模型 |
| DOI：10.11841/j.issn.1007-4333.2011.02.028 |
| 投稿时间：2010-07-21 |
| 基金项目:国家"十一五"科技支撑计划重点项目(2009BADA9B01); 中央级公益性科研院所基本科研业务费专项(2010-J-11) |
|
| Study on short-term forecasting methods and modelingof agro-product market price:Forecasting based on the time series models |
|
LI Gan-qiong, XU Shi-wei, LI Zhe-min, DONG Xiao-xia
|
| (Agricultural Information Institute,Chinese Academy of Agricultural Sciences/Key Lab ofDigital Agricultural Early Warning Technology,Ministry of Agriculture,Beijing 100081,China) |
| Abstract: |
| In order to improve the predictability of agro-product market price and to take measures to reduce price fluctuation,this study selected the wholesale price of tomatoes as an object and employed the five methods,the seasonal dummy variables,the Census X12 method,the moving average method,the Holt-Winters seasonal exponential smoothing method and the SARIMA,to establish short-term forecasting models.A combination forecasting model was established and the weights used in the model were calculated according to the single model prediction error.The results showed that the error of single model fluctuated greatly and the accuracy declined with longer forecast period.Mean absolute percent error (MAPE) of the five single models in forecasting evaluation for 2009 is about 10%,of which the Holt-Winters seasonal exponential smoothing model has the lowest MAPE of 6.81%.When forecast period is ahead of 3 months,the accuracy of the SARIMA model was the highest which was more than 95%.On the basis of empirical analysis,this study used the combination forecast method to predict the tomatoes market price in 2010. |
| Key words: agro-product market price forecasting model |
