在布朗(Brown)单一参数线性指数平滑法、ARIMA和GM(1,1)模型的基础上,利用我国农用化肥施用量数据,以绝对预测误差和达到最小建立组合预测模型。所得到的预测结果误差优于各个单一模型的单行预测,说明组合预测模型在时间序列数据的预测中更有可信度的优势。 Based on the Brownian single parameter linear exponential smoothing method, ARIMA and GM (1,1) models, we use the data of agricultural chemical fertilizer application in our country to establish the combined forecasting model with the absolute prediction error and minimum. The error of prediction result obtained is better than the single-line prediction of each single model, which shows that the combined forecasting model has more credibility advantage in the prediction of time-series data.