基于粒子群优化GRU的粮食产量预测  

Grain yield prediction based on Particle Swarm Optimization(PSO)and Gated Recurrent Unit(GRU)

作　　者：袁世一 YUAN Shiyi(Institute of Agricultural Information,Chinese Academy of Agricultural Sciences,Beijing 100081,China)

机构地区：[1]中国农业科学院农业信息研究所,北京100081

出　　处：《智能计算机与应用》2024年第8期109-114,共6页Intelligent Computer and Applications

基　　金：国家自然科学基金(62103418);中国农业科学院农业信息研究所基本科研业务费项目(JBYW-AII-2024-30)。

摘　　要：粮食产量预测是关系国计民生的重要问题,为了提高粮食产量预测的精度,本文提出了一种基于粒子群优化算法(PSO)—门控循环单元(Particle Swarm Optimization-Gate Recurrent Unit,PSO-GRU)的粮食产量预测模型。该模型利用PSO算法对GRU模型的关键参数进行寻优,解决了GRU易陷入局部最优解的问题;使用1949-2021年全国粮食产量数据作为数据源,利用PSO-GRU模型对全国粮食产量进行预测,并与GRU、PSO-RNN和PSO-LSTM等模型进行比较分析。实验结果表明,PSO-GRU模型能够充分挖掘时间序列特征,在全局收敛方面展现出显著优势,提高了粮食产量预测的准确性。Grain output forecasting is an important issue related to national transportation and people′s livelihood.In order to improve the accuracy of grain output forecasting,a new grain output forecasting model is proposed in this paper:Grain yield prediction model based on Particle Swarm Optimization-Gate Recurrent Unit(PSO-GRU).The model uses PSO algorithm to optimize the key parameters of GRU model,and solves the problem that GRU is easy to fall into the local optimal solution.In this study,the national grain production data from 1949 to 2021 was used as the data source,the PSO-GRU model was used to forecast the national grain production,and the comparison and analysis were made with other models(including GRU,PSO-RNN and PSO-LSTM).The experimental results show that the PSO-GRU model can fully explore the features of time series and show significant advantages in global convergence,thus improving the accuracy of grain yield prediction.

关 键 词：粮食产量预测 PSO-GRU 粒子群优化算法 门控循环单元 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]