基于多变量经验概率模型的中国粮食产量模拟预测分析

Short run forecasting on China's cereals production:A simulation by multivariate empirical probability distributions model

基于多变量经验概率模型(MVE)设定了A和B两个模拟方案对我国粮食产量进行模拟预测,并运用绝对误差均值(MAE)、均方根误差(RMSE)、误差相对比(MAPE)和希尔不等式系数(Theil IC)对模拟结果的预测精度进行了检验.结果表明,多变量经验概率模型的短期模拟预测精度较好,且模拟运用的历史数据时间间隔越短,模拟预测精度越高;根据2000-2008年的历史数据(B方案)和多变量经验概率模型模拟得到的2010年和2011年我国粮食产量的模拟均值分别为53288万吨和53887万吨,标准差分别为1967万吨和2001万吨.

The paper setted two scenarios of A and B based on a model of multivariate empirical probability distributions （MVE） to forecast China＇s cereals production in a short run. Four measures of the mean absolute error （MAE）, root mean square error （RMSE）, mean absolute percentage error （MAPE） and Theil＇s inequality coefficient （Theil IC） were used to test ex post simulation fit. Result indicates that the MVE model＇s short-term prediction accuracy is better simulated, and if the deadline of historical data is closer to the starting time, the statistic fit is better. The paper also finds that according to historical data from 2000 to 2008 （B scenario）, the simulated mean values of China＇s cereals production in 2010 and 2011 are 532.88 million tons and 538.87 million tons, and the standard deviation values are 19.67 million tons and 20.01 million tons, respectively.