湖北农业科学 ›› 2025, Vol. 64 ›› Issue (5): 155-159.doi: 10.14088/j.cnki.issn0439-8114.2025.05.024

• 信息工程 • 上一篇    下一篇

基于Stacking集成算法的中国南方地区粮食产量预测

马滇璟1, 赵家松1, 严伟榆1, 段光俊1, 刘振洋2, 吴绍天1   

  1. 1.云南农业大学大数据学院,昆明 650201;
    2.昆明城市学院数据科学与工程学院,昆明 650032
  • 收稿日期:2024-11-18 出版日期:2025-05-25 发布日期:2025-06-11
  • 通讯作者: 赵家松(1975-),男,云南昆明人,副教授,博士,主要从事数据挖掘、人工智能与农业应用等工作,(电话)13033399944(电子信箱)zhaojiasong@ynau.edu.cn。
  • 作者简介:马滇璟(2000-),女,云南红河人,硕士,主要从事粮食产量预测研究,(电话)15287300714(电子信箱)1838437565@qq.com
  • 基金资助:
    云南省农业基础研究联合专项基金资助项目(202301BD070001-202); 云南农业大学博士科研启动基金资助项目(A2032002507)

Grain yield prediction in southern China based on Stacking ensemble algorithm

MA Dian-jing1, ZHAO Jia-song1, YAN Wei-yu1, DUAN Guang-jun1, LIU Zhen-yang2, WU Shao-tian1   

  1. 1. School of Big Data, Yunnan Agricultural University, Kunming 650201, China;
    2. School of Data Science and Engineering, Kunming City College, Kunming 650032, China
  • Received:2024-11-18 Published:2025-05-25 Online:2025-06-11

摘要: 基于中国南方地区1998—2022年安徽省、湖北省、湖南省、江苏省和四川省的粮食产量及11个维度的相关因素数据,构建基于Stacking集成算法的BP-SVR-Stacking粮食产量预测模型,并将其与BP神经网络模型和SVR模型进行对比分析。结果表明,BP-SVR-Stacking模型的平均绝对误差(MAE)和平均绝对百分比误差(MAPE)均低于BP神经网络模型和SVR模型,说明BP-SVR-Stacking模型的预测能力优于单一的机器学习模型。相较于BP神经网络模型和SVR模型,BP-SVR-Stacking模型的决定系数(R2)分别提高了0.124和0.122,说明BP-SVR-Stacking模型具有良好的拟合能力和预测性能。

关键词: Stacking集成算法, 粮食产量, 中国南方, 预测

Abstract: Based on the grain yield data and 11-dimensional relevant factors from Anhui, Hubei, Hunan, Jiangsu, and Sichuan provinces in southern China between 1998 and 2022,the BP-SVR-Stacking grain yield prediction model based on the Stacking ensemble algorithm was developed and comparatively analyzed with the BP neural network model and SVR model. The results indicated that the mean absolute error (MAE) and mean absolute percentage error (MAPE) of the BP-SVR-Stacking model were significantly lower than those of the BP neural network model and SVR model,which demonstrated the superior prediction capability of the BP-SVR-Stacking model over single machine learning models. Compared with the BP neural network model and SVR model, the coefficient of determination (R2) of the BP-SVR-Stacking model increased by 0.124 and 0.122 respectively,suggesting that the BP-SVR-Stacking model possessed excellent fitting capability and prediction performance.

Key words: Stacking ensemble algorithm, grain yield, southern China, prediction

中图分类号: