基于ARIMA-GM组合模型的农作物播种面积预测——以吉林省为例

Crop sowing area forecasting based on ARIMA-GM combined model——Taking Jilin Province as an example

农作物种植面积预测能够有效帮助生产者提前掌握未来农业的生产状况。为准确预测农作物播种面积变化情况,寻求一种新型的组合模型权重计算方法。以吉林省历年作物播种面积为原始数据,建立GM(1,1)与ARIMA时间序列组合预测模型。综合使用平均绝对百分数误差、残差平方和最小法及最小二乘法确定组合模型的权系数,在此基础上将三种权系数组合优化后得到最终权重,两种模型权重分别为0.44976和0.55024。预测结果表明:该组合权重与单一权重相比,可将组合模型的预测精度提高约0.4%。

Crop acreage forecasting can effectively help producers to grasp the future production status of agriculture in advance.In order to accurately predict the change of crop sown area,a new combined model weight calculation method is sought.Using the historical crop sown area of Jilin Province as the original data,a forecasting model of combining GM(1,1)and ARIMA time series is built.Combining mean absolute percentage error,residual sum-of-squares minimization and least squares to determine the weighting factors for the portfolio model,the final weights are obtained by optimizing and combining the three weighting factors,two model weights are 0.44976 and 0.55024 respectively.The prediction results show that the combination weight can improve the prediction accuracy of the combined model by approximately 0.4%compared with single weight.