基于鹈鹕算法(POA)优化卷积神经网络-长短期记忆网络(CNN-LSTM)回归预测，POA-CNN-LSTM多输入单输出模型

%% CNN-LSTM多变量回归预测 %% 清空环境变量 warning off % 关闭报警信息 close all % 关闭开启的图窗 clear % 清空变量 clc % 清空命令行 % restoredefaultpath tic %% 加载数据与数据集划分 %% 导入数据 P_train = xlsread('data','training set','B2：G191')'; T_train= xlsread('data','training set','H2：H191')'; % 测试集——44个样本 P_test=xlsread('data','test set','B2:G45')'; T_test=xlsread('data','test set','H2:H45')'; f_ = size(P_train, 1); % 输入特征维度 %% 划分训练集和测试集 M = size(P_train, 2); N = size(P_test, 2); %% 数据归一化 [p_train, ps_input] = mapminmax(P_train, 0, 1); p_test = mapminmax('apply', P_test, ps_input); [t_train, ps_output] = mapminmax(T_train, 0, 1); t_test = mapminmax('apply', T_test, ps_output); %% 创建元胞或向量，长度为训练集大小； XrTrain = cell(size(p_train,2),1); YrTrain = zeros(size(t_train,2),1); for i=1:size(p_train,2) XrTrain{i,1} = p_train(:,i); YrTrain(i,1) = t_train(:,i); end % 创建元胞或向量，长度为测试集大小； XrTest = cell(size(p_test,2),1); YrTest = zeros(size(t_test ,2),1); for i=1:size(p_test,2) XrTest{i,1} = p_test(:,i); YrTest(i,1) = t_test (:,i); end %% 优化算法参数设置 SearchAgents_no = 8; % 数量 Max_iteration = 5; % 最大迭代次数 dim = 3; % 优化参数个数 lb = [1e-3,10 1e-4]; % 参数取值下界(学习率，隐藏层节点，正则化系数) ub = [1e-2, 30,1e-1]; % 参数取值上界(学习率，隐藏层节点，正则化系数) fitness = @(x)fical(x,XrTrain,YrTrain,ps_output,T_train,f_); [Best_score,Best_pos,curve]=POA(SearchAgents_no,Max_iteration,lb ,ub,dim,fitness) best_hd = round(Best_pos(1, 2)); % 最佳隐藏层节点数 best_lr= Best_pos(1, 1);% 最佳初始学习率 best_l2 = Best_pos(1, 3);% 最佳L2正则化系数 Best_pos(1, 2) = round(Best_pos(1, 2)); %% 创建混合CNN-LSTM网络架构 % 输入特征维度 numFeatures = f_; % 输出特征维度 numResponses = 1; FiltZise = 10; % 创建"CNN-LSTM"模型 layers = [... % 输入特征 sequenceInputLayer([numFeatures 1 1],'Name','input') sequenceFoldingLayer('Name','fold') % CNN特征提取 convolution2dLayer([FiltZise 1],32,'Padding','same','WeightsInitializer','he','Name','conv','DilationFactor',1); batchNormalizationLayer('Name','bn') eluLayer('Name','elu') averagePooling2dLayer(1,'Stride',FiltZise,'Name','pool1') % 展开层 sequenceUnfoldingLayer('Name','unfold') % 平滑层 flattenLayer('Name','flatten') % LSTM特征学习 lstmLayer(50,'Name','lstm1','RecurrentWeightsInitializer','He','InputWeightsInitializer','He') % LSTM输出 lstmLayer(best_hd,'OutputMode',"last",'Name','bil4','RecurrentWeightsInitializer','He','InputWeightsInitializer','He') dropoutLayer(0.25,'Name','drop3') % 全连接层 fullyConnectedLayer(numResponses,'Name','fc') regressionLayer('Name','output') ]; layers = layerGraph(layers); layers = connectLayers(layers,'fold/miniBatchSize','unfold/miniBatchSize'); %% CNNLSTM训练选项 % 批处理样本 MiniBatchSize =128; % 最大迭代次数 MaxEpochs = 500; options = trainingOptions( 'adam', ... 'MaxEpochs',500, ... 'GradientThreshold',1, ... 'InitialLearnRate',best_lr, ... 'LearnRateSchedule','piecewise', ... 'LearnRateDropPeriod',400, ... 'LearnRateDropFactor',0.2, ... 'L2Regularization',best_l2,... 'Verbose',false, ... 'Plots','training-progress'); %% 训练混合网络 net = trainNetwork(XrTrain,YrTrain,layers,options); %% 训练集误差评价 % 预测 YPredtr = predict(net,XrTrain,"MiniBatchSize",numFeatures); % 结果 t_sim1 =double(YPredtr'); % 反归一化 CNNLSTMoutput_tra=mapminmax('reverse',t_sim1,ps_output); T_sim1=double(CNNLSTMoutput_tra); %% 测试集误差评价 % 预测 YPred = predict(net,XrTest,"MiniBatchSize",numFeatures); % 结果 t_sim2=double(YPred'); % 反归一化 CNNLSTMoutput_test=mapminmax('reverse',t_sim2,ps_output); T_sim2=double(CNNLSTMoutput_test); toc %% 训练集误差评价 figure plot(1 : length(curve), curve,'linewidth',1.5); title('POA-CNN-LSTM', 'FontSize', 10); xlabel('迭代次数', 'FontSize', 10); ylabel('适应度值mse', 'FontSize', 10); grid off %% 测试集结果 figure; plotregression(T_test,T_sim2,['回归图']); figure; ploterrhist(T_test-T_sim2,['误差直方图']); %% 均方根误差 RMSE error1 = sqrt(sum((T_sim1 - T_train).^2)./M); error2 = sqrt(sum((T_test - T_sim2).^2)./N); %% %决定系数 R1 = 1 - norm(T_train - T_sim1)^2 / norm(T_train - mean(T_train))^2; R2 = 1 - norm(T_test - T_sim2)^2 / norm(T_test - mean(T_test ))^2; %% %均方误差 MSE mse1 = sum((T_sim1 - T_train).^2)./M; mse2 = sum((T_sim2 - T_test).^2)./N; %% %RPD 剩余预测残差 SE1=std(T_sim1-T_train); RPD1=std(T_train)/SE1; SE=std(T_sim2-T_test); RPD2=std(T_test)/SE; %% 平均绝对误差MAE MAE1 = mean(abs(T_train - T_sim1)); MAE2 = mean(abs(T_test - T_sim2)); %% 平均绝对百分比误差MAPE MAPE1 = mean(abs((T_train - T_sim1)./T_train)); MAPE2 = mean(abs((T_test - T_sim2)./T_test)); %% 训练集绘图 figure %plot(1:M,T_train,'r-*',1:M,T_sim1,'b-o','LineWidth',1) plot(1:M,T_train,'r-*',1:M,T_sim1,'b-o','LineWidth',1.5) legend('真实值','POA-CNN-LSTM预测值') xlabel('预测样本') ylabel('预测结果') string={'训练集预测结果对比';['(R^2 =' num2str(R1) ' RMSE= ' num2str(error1) ' MSE= ' num2str(mse1) ' RPD= ' num2str(RPD1) ')' ]}; title(string) %% 预测集绘图 figure plot(1:N,T_test,'r-*',1:N,T_sim2,'b-o','LineWidth',1.5) legend('真实值','POA-CNN-LSTM预测值') xlabel('预测样本') ylabel('预测结果') string={'测试集预测结果对比';['(R^2 =' num2str(R2) ' RMSE= ' num2str(error2) ' MSE= ' num2str(mse2) ' RPD= ' num2str(RPD2) ')']}; title(string) %% 测试集误差图 figure ERROR3=T_test-T_sim2; plot(T_test-T_sim2,'b-*','LineWidth',1.5) xlabel('测试集样本编号') ylabel('预测误差') title('测试集预测误差') grid on; legend('预测输出误差') %% 绘制线性拟合图 %% 训练集拟合效果图 figure plot(T_train,T_sim1,'*r'); xlabel('真实值') ylabel('预测值') string = {'训练集效果图';['R^2_c=' num2str(R1) ' RMSEC=' num2str(error1) ]}; title(string) hold on ;h=lsline; set(h,'LineWidth',1,'LineStyle','-','Color',[1 0 1]) %% 预测集拟合效果图 figure plot(T_test,T_sim2,'ob'); xlabel('真实值') ylabel('预测值') string1 = {'测试集效果图';['R^2_p=' num2str(R2) ' RMSEP=' num2str(error2) ]}; title(string1) hold on ;h=lsline(); set(h,'LineWidth',1,'LineStyle','-','Color',[1 0 1]) %% 求平均 R3=(R1+R2)./2; error3=(error1+error2)./2; %% 总数据线性预测拟合图 tsim=[T_sim1,T_sim2]'; S=[T_train,T_test]'; figure plot(S,tsim,'ob'); xlabel('真实值') ylabel('预测值') string1 = {'所有样本拟合预测图';['R^2_p=' num2str(R3) ' RMSEP=' num2str(error3) ]}; title(string1) hold on ;h=lsline(); set(h,'LineWidth',1,'LineStyle','-','Color',[1 0 1]) %% 打印出评价指标 disp(['-----------------------误差计算--------------------------']) disp(['评价结果如下所示：']) disp(['平均绝对误差MAE为：',num2str(MAE2)]) disp(['均方误差MSE为： ',num2str(mse2)]) disp(['均方根误差RMSEP为： ',num2str(error2)]) disp(['决定系数R^2为： ',num2str(R2)]) disp(['剩余预测残差RPD为： ',num2str(RPD2)]) disp(['平均绝对百分比误差MAPE为： ',num2str(MAPE2)]) grid disp(['学习率，隐藏层节点，正则化系数',num2str(Best_pos)])