基于CNN-BiLSTM-Attention模型的胡麻产量预测

打开文本图片集

中图分类号：S565.9 文献标识码：A 文章编号： 1000-4440（2025）07-1342-（

Flax yield prediction based on CNN-BiLSTM-Attention model

LI Xingyu'，LI Yue 1，2 ， GAO Yuhong2，3（1.Colfdlsi；ebHabtats，Uesi；fes

Abstract：This study proposed a deep learning-based model integrating Convolutional Neural Network （CNN），BidirectionalLongShort-TermMemory（BiLSTM）andatentionmechanismtopredictflaxyield.Themodelcombinedthe spatialfeature extractioncapabilityofCNN，the temporal dynamicmodelingabilityof BiLSTM，and the featureadaptive weighting functionof the Atentionmechanism.The model was trained using climatedata，vegetation indices，andyielddata during 2000-2020.Experimentalresultsshowed that theCNN-BiLSTM-Attntionmodel significantlyoutperformedtraditional models in prediction accuracy，with a root mean square error （ RMSE ）of 316.98kg/hm2 and a coefficient of determination （ R2 ）of 0.83.The model maintained good stabilityand high accuracy under interannual climate change conditions.This studyprovides technical supportforflaxyieldprediction，and itsmodulardesign frameworkcanalsobeextended tohe growth monitoring and yield prediction of other crops.

Key words：flax；yield prediction；deep learning；Convolutional Neural Network；Bidirectional Long Short-Term Memory model

胡麻（油用亚麻）是亚麻科（Liancease）亚麻属（Linum）的重要油料作物[1]，其籽粒富含不饱和脂肪酸（亚麻酸、亚油酸、油酸）、木酚素和亚麻胶等活性物质，具有降血脂、降血压、降血糖以及预防心脑血管疾病等功能[2]。（剩余11298字）