SSA-GRU、GRU麻雀算法优化门控循环单元时间序列预测对比（Matlab完整程序和数据）

%% 基于麻雀算法优化的GRU预测算法 clear;close all; clc rng('default') %% 仿真数据，正弦信号 data = sin(0:1:200)'; %序列的前 90% 用于训练，后 10% 用于测试 numTimeStepsTrain = floor(0.9*numel(data)); dataTrain = data(1:numTimeStepsTrain+1)'; dataTest = data(numTimeStepsTrain+1:end)'; %数据预处理，将训练数据标准化为具有零均值和单位方差。 mu = mean(dataTrain); sig = std(dataTrain); dataTrainStandardized = (dataTrain - mu) / sig; %输入GRU的时间序列交替一个时间步 XTrain = dataTrainStandardized(1:end-1); YTrain = dataTrainStandardized(2:end); XTest = dataTest(1:end-1); YTest = dataTest(2:end); %% %创建GRU回归网络，指定GRU层的隐含单元个数96*3 %序列预测，因此，输入一维，输出一维 numFeatures = 1; numResponses = 1; %% 定义麻雀优化参数 pop=5; %种群数量 Max_iteration=10; % 设定最大迭代次数 dim = 4;%维度，即GRU网路包含的隐藏单元数目,最大训练周期,初始学习率,L2参数 lb = [2,2,10E-3,10E-3];%下边界 ub = [50,100,1,1];%上边界 fobj = @(x) fun(x,numFeatures,numResponses,XTrain,YTrain,XTest,YTest); [Best_pos,Best_score,SSA_curve,netSSA]=SSA(pop,Max_iteration,lb,ub,dim,fobj); %开始优化 figure plot(SSA_curve,'linewidth',1.5); grid on xlabel('迭代次数') ylabel('适应度函数') title('SSA-GRU收敛曲线') %训练集测试 PredictTrainSSA = predict(netSSA,XTrain, 'ExecutionEnvironment','cpu'); %测试集测试 PredictTestSSA = predict(netSSA,XTest, 'ExecutionEnvironment','cpu'); %训练集mse mseTrainSSA= mse(YTrain-PredictTrainSSA); %测试集mse mseTestSSA = mse(YTest-PredictTestSSA); %% 基础GRU测试 numHiddenUnits = 100; layers = [ ... sequenceInputLayer(numFeatures) gruLayer(numHiddenUnits) fullyConnectedLayer(numResponses) regressionLayer]; %指定训练选项 options = trainingOptions('adam', ... 'MaxEpochs',500, ... 'ExecutionEnvironment' ,'cpu',... 'GradientThreshold',1, ... 'InitialLearnRate',0.001, ... 'L2Regularization',0.0001,... 'Verbose',0); %训练GRU net = trainNetwork(XTrain,YTrain,layers,options); %训练集测试 PredictTrain = predict(net,XTrain, 'ExecutionEnvironment','cpu'); %测试集测试 PredictTest = predict(net,XTest, 'ExecutionEnvironment','cpu'); %训练集mse mseTrain = mse(YTrain-PredictTrain); %测试集mse mseTest = mse(YTest-PredictTest); disp('-------------------------------------------------------------') disp('SSA-GRU优化得到的最优参数为：') disp(['SSA-GRU优化得到的隐藏单元数目为：',num2str(round(Best_pos(1)))]); disp(['SSA-GRU优化得到的最大训练周期为：',num2str(round(Best_pos(2)))]); disp(['SSA-GRU优化得到的InitialLearnRate为：',num2str((Best_pos(3)))]); disp(['SSA-GRU优化得到的L2Regularization为：',num2str((Best_pos(4)))]); disp('-------------------------------------------------------------') disp('SSA-GRU结果：') disp(['SSA-GRU训练集MSE:',num2str(mseTrainSSA)]); disp(['SSA-GRU测试集MSE:',num2str(mseTestSSA)]); disp('GRU结果：') disp(['GRU训练集MSE:',num2str(mseTrain)]); disp(['GRU测试集MSE:',num2str(mseTest)]); %% 训练集结果绘图 figure errors=YTrain-PredictTrain; errorsSSA=YTrain-PredictTrainSSA; MSE=mean(errors.^2); RMSE=sqrt(MSE); MSESSA=mean(errorsSSA.^2); RMSESSA=sqrt(MSESSA); error_mean=mean(errors); error_std=std(errors); error_meanSSA=mean(errorsSSA); error_stdSSA=std(errorsSSA); subplot(2,2,[1 2]); plot(YTrain,'k'); hold on; plot(PredictTrain,'b'); plot(PredictTrainSSA,'r'); legend('Target','GRU','SSA_GRU'); title('训练集结果'); xlabel('Sample Index'); grid on; subplot(2,2,3); plot(errors); hold on plot(errorsSSA); legend('GRU-Error','SSA-GRU-Eoor'); title(['MSE = ' num2str(MSESSA) ', RMSE = ' num2str(RMSESSA),'MSESSAS = ']); grid on; subplot(2,2,4); histfit(errorsSSA, 50); title(['Error Mean = ' num2str(error_mean) ', Error St.D. = ' num2str(error_std)]); %% 测试集结果绘图 figure errors=YTest-PredictTest; errorsSSA=YTest-PredictTestSSA; MSE=mean(errors.^2); RMSE=sqrt(MSE); MSESSA=mean(errorsSSA.^2); RMSESSA=sqrt(MSESSA); error_mean=mean(errors); error_std=std(errors); error_meanSSA=mean(errorsSSA); error_stdSSA=std(errorsSSA); subplot(2,2,[1 2]); plot(YTest,'k'); hold on; plot(PredictTest,'b'); plot(PredictTestSSA,'r'); legend('Target','GRU','SSA-GRU'); title('测试集结果'); xlabel('Sample Index'); grid on; subplot(2,2,3); plot(errors); hold on plot(errorsSSA); legend('GRU-Error','SSA-GRU-Eoor'); title(['MSE = ' num2str(MSESSA) ', RMSE = ' num2str(RMSESSA),'MSESSAS = ']); grid on; subplot(2,2,4); histfit(errorsSSA, 50); title(['Error Mean = ' num2str(error_mean) ', Error St.D. = ' num2str(error_std)]);