基于IPSO-BP模型的粮食产量预测方法研究

针对粮食产量预测的复杂性,以基本微粒群算法(PSO)为基础,引入繁殖和变异机制,提出一种改进的微粒群算法(IPSO)优化BP神经网络的连接权值和阈值。综合考虑影响粮食产量的有关因素,构建出IPSO-BP的粮食产量预测模型,并以江苏省1978—2018年的粮食产量及影响其产量的10个因素作为数据集进行仿真试验。试验结果表明,与基本PSO-BP神经网络模型及BP神经网络模型相比,基于IPSO-BP神经网络模型获得的2016年、2017年、2018年粮食产量预测结果最优相对误差分别为0.24%、0.25%、0.06%,平均相对误差0.76%、0.67%、0.38%,该模型的预测精度及稳定性大幅提高。

Research on prediction method of grain yield based on IPSO-BP model

In view of the complexity of grain yield prediction,based on the basic particle swarm optimization(PSO)algorithm,an improved particle swarm optimization(IPSO)algorithm combines reproduction and variation mechanism is proposed,which is used to optimize weights and thresholds of BPNN.Considering the factors that affect the grain yield,the prediction model of IPSO-BP is established.The grain yield of Jiangsu Province from 1978 to 2018 and ten factors affecting the yield are used as data sets for simulation experiments.The experimental results show that compared with the basic PSO-BP neural network model and BP neural network model,the optimal relative errors of grain yield forecast results using IPSO-BP neural network model in 2016,2017 and 2018 are 0.24%,0.25%and 0.06%respectively,the average relative errors of 0.76%,0.67%and 0.38%.Therefore,the prediction accuracy and stability of the model are greatly improved.