甘蔗糖业2025，Vol.54Issue(2)：62-68,7.DOI:10.3969/j.issn.1005-9695.2025.02.007
基于多元数据的甘蔗产量预报
Sugarcane Yield Forecast based on Multivariate Data
摘要
Abstract
To improve the level of intelligent prediction of sugarcane yield,this study took the annual yield of sugarcane in 14 cities of Guangxi from 2021 to 2023 as the research object.Combining with the daily average value data of meteorological elements,all elements were reduced to one dimension through the principal component analysis method,and the predicted values of meteorological elements and sugarcane yield under four different machine learning algorithms were compared.The results showed that the yield prediction error of the random forest algorithm was the smallest,which was 0.082 t,and the second lowest error value was 0.091 t for support vector regression;the error of K-nearest neighbor regression was 0.098 t when the K value was 4;the multiple linear regression algorithm performed the worst in the fitting process,with an error of 0.921 t.The most important meteorological and soil elements in the fitting process were relative humidity,air temperature,and soil moisture,respectively.The research results can provide scientific support for the intelligent prediction of sugarcane yield in Guangxi.关键词
主成分分析法降维/机器学习/甘蔗产量/气象/土壤Key words
Principal component analysis dimensionality reduction/Machine learning/Sugarcane yield/Meteorological/Soil分类
农业科学引用本文复制引用
陶伟,粟华林,成振华,王玮,吕善行,周坤论,朱鱼扬,李强..基于多元数据的甘蔗产量预报[J].甘蔗糖业,2025,54(2):62-68,7.基金项目
中国气象局科技项目"揭榜挂帅""基于天空地一体化融合数据的精细化甘蔗农情灾情评估预警和产量预报技术与数算一体化平台研发"(CMAJBGS202322) （CMAJBGS202322）
广西气象局"指令性项目""广西'天衡天衍'系统本地化研究"(桂气科 2024ZL04) （桂气科 2024ZL04）
