基于KPCA-SSA-BP的农业气象灾害预测

Agricultural meteorological disaster prediction based on KPCA-SSA-BP

农业气象灾害对农业发展有很大阻碍,为优化农业气象灾害预测的估算模型,本研究以山东省作为研究区域,利用核主成分分析(KPCA)对影响因子进行降维,以传统反向传播(BP)神经网络模型为基础,基于麻雀搜索算法(SSA)、粒子群算法(PSO)、遗传算法(GA)3种优化算法,构建了SSA-BP、PSO-BP、GA-BP 3种优化模型。结果表明,在旱灾受灾率的模型评价指标对比中,发现与传统BP神经网络模型相比,SSA-BP、PSO-BP、GA-BP神经网络模型的均方根误差(RMSE)分别下降23.55%、12.28%和17.74%;在洪灾受灾率的模型评价对比中,发现与传统BP神经网络模型相比,SSA-BP、PSO-BP、GA-BP神经网络模型的RMSE分别下降了29.96%、9.49%和13.88%。说明SSA-BP神经网络模型对旱灾受灾率、洪灾受灾率的预测效果优于传统BP神经网络模型以及PSO-BP、GA-BP优化的神经网络模型。

Agrometeorological disasters are great obstacles to agricultural development.In order to optimize the estimation model for agrometeorological disaster prediction,this study took Shandong province as the study area,and used the kernel principal component analysis(KPCA)to downscale the influencing factors.Based on the traditional backpropagation(BP)neural network model,the sparrow search algorithm(SSA),particle swarm algorithm(PSO),and genetic algorithm(GA)optimization algorithms were used to construct three optimization models,SSA-BP,PSO-BP and GA-BP.The results showed that in the prediction of drought disaster rate,compared with the traditional BP neural network model,the root mean square error(RMSE)of SSA-BP,PSO-BP and GA-BP neural network models decreased by 23.55%,12.28%and 17.74%respectively.In the prediction of flood disaster rate,compared with the traditional BP neural network model,the RMSE of SSA-BP,PSO-BP and GA-BP neural network models decreased by 29.96%,9.49%and 13.88%respectively.These results indicated that the SSA-BP model was better than the traditional BP neural network model,PSO-BP neural network model and GA-BP neural network model in predicting the damage rate of drought and flood.