【LSTM回归预测】基于双向长短时记忆BiLSTM数据回归预测（多输入单输出）含Matlab源码.zip

%环境准备 %清理工作区间及命令窗口 clc;clear; warning off; %导入数据 data=xlsread('data'); %准备输入和输出训练数据 input =data(:,1:4)'; output=data(:,5)'; nwhole =size(data,1); train_ratio=0.95; ntrain=round(nwhole*train_ratio); ntest =nwhole-ntrain; input_train =input(:,1:ntrain); output_train=output(:,1:ntrain); input_test =input(:, ntrain+1:ntrain+ntest); output_test=output(:,ntrain+1:ntrain+ntest); %归一化（全部特征 均归一化） [inputn_train,inputps] =mapminmax(input_train); [outputn_train,outputps]=mapminmax(output_train); inputn_test =mapminmax('apply',input_test,inputps); outputn_test=mapminmax('apply',output_test,outputps); %特征维度确定 %数据输入x的特征维度 inputSize = size(inputn_train,1); %数据输出y的维度 outputSize = size(outputn_train,1); %可以增加对照实验确定 numhidden_units=200; %BiLSTM 层参数设置 %输入层设置 sequenceInputLayer %学习层设置 bilstmLayer %全连接层设置 fullyConnectedLayer %epoch后学习率更新 %% bilstm layers = [ ... sequenceInputLayer(inputSize) bilstmLayer(numhidden_units) dropoutLayer(0.2) fullyConnectedLayer(outputSize) regressionLayer('name','out')]; %% trainoption(bilstm) opts = trainingOptions('adam', ... 'MaxEpochs',200, ... 'GradientThreshold',1,... 'ExecutionEnvironment','cpu',... 'InitialLearnRate',0.005, ... 'LearnRateSchedule','piecewise', ... 'LearnRateDropPeriod',100, ... 'LearnRateDropFactor',0.8, ... 'Verbose',0, ... 'Plots','training-progress'... ); %BiLSTM网络训练 tic BiLSTMnet = trainNetwork(inputn_train ,outputn_train ,layers,opts); toc; [BiLSTMnet,BiLSTMoutputr_train]= predictAndUpdateState(BiLSTMnet,inputn_train); BiLSTMoutput_train = mapminmax('reverse',BiLSTMoutputr_train,outputps); %BiLSTM网络测试 [BiLSTMnet,BiLSTMoutputr_test] = predictAndUpdateState(BiLSTMnet,inputn_test); %网络输出反归一化 BiLSTMoutput_test= mapminmax('reverse',BiLSTMoutputr_test,outputps); %BiLSTM数据评价 error_test=BiLSTMoutput_test'-output_test'; pererror_test=error_test./output_test'; error=error_test'; pererror=pererror_test'; avererror=sum(abs(error))/(ntest); averpererror=sum(abs(pererror))/(ntest); RMSE = sqrt(mean((error).^2)); disp('BiLSTM网络预测绝对平均误差MAE'); disp(avererror); disp('BiLSTM网络预测平均绝对误差百分比MAPE'); disp(averpererror) disp('BiLSTM网络预测均方根误差RMSE') disp(RMSE) % 数据可视化分析 %测试数据 figure() plot(BiLSTMoutput_test,'r-.') hold on plot(output_test,'k--') legend( '预测测试数据','实际分析数据','Location','NorthWest','FontName','仿宋'); title('BiLSTM网络模型结果及真实值','fontsize',15,'FontName','仿宋') xlabel('样本','fontsize',10,'FontName','仿宋'); ylabel('数值','fontsize',10,'FontName','仿宋'); %------------------------------------------------------------------------------------- figure() stairs(pererror_test,'-.','Color',[255 50 0]./255,'linewidth',1) legend('BiLSTM网络测试相对误差','Location','NorthEast','FontName','仿宋') title('BiLSTM网络预测相对误差','fontsize',10,'FontName','仿宋') ylabel('误差','fontsize',10,'FontName','仿宋') xlabel('样本','fontsize',10,'FontName','仿宋') %-------------------------------------------------------------------------------------