时序预测 - MATLAB实现Bayes（贝叶斯）优化LSTM(长短期记忆神经网络)时间序列预测（完整源码和数据）

clc; clear; close all; %% 变量初始化 opt.Delays = 1:30; %数据标准化: 'None' 'Data Standardization' 'Data Normalization' opt.dataPreprocessMode = 'Data Standardization'; opt.learningMethod = 'LSTM'; %数据划分 opt.trPercentage = 0.80; %% 参数定义 %最大迭代次数 opt.maxEpochs = 10; %批处理大小 opt.miniBatchSize = 16; %执行环境： 'cpu' 'gpu' 'auto' opt.executionEnvironment = 'cpu'; %优化算法： 'sgdm' 'rmsprop' 'adam' opt.LR = 'adam'; %训练进程： 'training-progress' 'none' opt.trainingProgress = 'none'; %% 双向长短期参数设定 %逻辑参数控制是否双向 opt.isUseBiLSTMLayer = true; opt.isUseDropoutLayer = true; %权重丢失参数设定,防止过拟合 opt.DropoutValue = 0.5; % 优化参数设定 opt.optimVars = [ optimizableVariable('NumOfLayer',[1 4],'Type','integer') optimizableVariable('NumOfUnits',[50 200],'Type','integer') optimizableVariable('isUseBiLSTMLayer',[1 2],'Type','integer') optimizableVariable('InitialLearnRate',[1e-2 1],'Transform','log') optimizableVariable('L2Regularization',[1e-10 1e-2],'Transform','log')]; opt.isUseOptimizer = true; opt.MaxOptimizationTime = 14*60*60; opt.MaxItrationNumber = 10; opt.isDispOptimizationLog = true; %参数保存设定 opt.isSaveOptimizedValue = false; opt.isSaveBestOptimizedValue = true; %% 数据导入 data.CompleteData = readtable('data.xlsx','Sheet',1); data.seriesdataHeder = data.CompleteData.Properties.VariableNames(1,:); data.seriesdata = table2array(data.CompleteData(:,:)); disp('数据导入成功'); data.isDataRead = true; figure('Name','InputData','NumberTitle','off'); plot(data.seriesdata,'--','Color',[0 0 180]./255,'linewidth',2,'Markersize',4,'MarkerFaceColor',[0 0 180]./255); grid minor; title({['Mean = ' num2str(mean(data.seriesdata)) ', STD = ' num2str(std(data.seriesdata)) ];}); if strcmpi(opt.dataPreprocessMode,'None') data.x = data.seriesdata; elseif strcmpi(opt.dataPreprocessMode,'Data Normalization') for i=1:size(data.seriesdata,2) data.x(:,i) = (data.seriesdata(:,i) -min(data.seriesdata(:,i)))./ (max(data.seriesdata(:,i))-min(data.seriesdata(:,i))); end figure('Name','NormilizedInputData','NumberTitle','off'); plot(data.x,'--','Color',[255 0 0]./255,'linewidth',2,'Markersize',4,'MarkerFaceColor',[255 0 0]./255); grid minor; title({['Mean = ' num2str(mean(data.x)) ', STD = ' num2str(std(data.x)) ];}); elseif strcmpi(opt.dataPreprocessMode,'Data Standardization') for i=1:size(data.seriesdata,2) x.mu(1,i) = mean(data.seriesdata(:,i),'omitnan'); x.sig(1,i) = std (data.seriesdata(:,i),'omitnan'); data.x(:,i) = (data.seriesdata(:,i) - x.mu(1,i))./ x.sig(1,i); end figure('Name','NormilizedInputData','NumberTitle','off'); plot(data.x,'--','Color',[255 0 0]./255,'linewidth',2,'Markersize',4,'MarkerFaceColor',[255 0 0]./255); grid minor; title({['Mean = ' num2str(mean(data.x)) ', STD = ' num2str(std(data.x)) ];}); end %% 数据准备 % LSTM数据准备 % 网络时间延迟设定 Delays = opt.Delays; x = data.x'; T = size(x,2); MaxDelay = max(Delays); Range = MaxDelay+1:T; X= []; for d = Delays X=[X; x(:,Range-d)]; end Y = x(:,Range); data.X = X; data.Y = Y; % 数据划分 data.XTr = []; data.YTr = []; data.XTs = []; data.YTs = []; numTrSample = round(opt.trPercentage*size(data.X,2)); data.XTr = data.X(:,1:numTrSample); data.YTr = data.Y(:,1:numTrSample); data.XTs = data.X(:,numTrSample+1:end); data.YTs = data.Y(:,numTrSample+1:end); disp(['Time Series data divided to ' num2str(opt.trPercentage*100) '% Train data and ' num2str((1-opt.trPercentage)*100) '% Test data']); % 数据规范化 for i=1:size(data.XTr,2) XTr{i,1} = data.XTr(:,i); YTr(i,1) = data.YTr(:,i); end for i=1:size(data.XTs,2) XTs{i,1} = data.XTs(:,i); YTs(i,1) = data.YTs(:,i); end data.XTr = XTr; data.YTr = YTr; data.XTs = XTs; data.YTs = YTs; data.XVl = XTs; data.YVl = YTs; disp('输入数据时间序列适应LSTM'); %% 基于贝叶斯寻找最佳LSTM参数 % 超参数运行 if opt.isDispOptimizationLog isLog = 2; else isLog = 0; end if opt.isUseOptimizer opt.ObjFcn = ObjFcn(opt,data); BayesObject = bayesopt(opt.ObjFcn,opt.optimVars, ... 'MaxTime',opt.MaxOptimizationTime, ... 'IsObjectiveDeterministic',false, ... 'MaxObjectiveEvaluations',opt.MaxItrationNumber,... 'Verbose',isLog,... 'UseParallel',false); end %% 数据评价 [opt,data] = EvaluationData(opt,data); %% 本地子函数 % 目标函数 function ObjFcnn = ObjFcn(opt,data) ObjFcnn = @CostFunction; function [valError,cons,fileName] = CostFunction(optVars) inputSize = size(data.X,1); numResponses = 1; dropoutVal = 0.5; if optVars.isUseBiLSTMLayer == 2 optVars.isUseBiLSTMLayer = 0; end % if dropout layer is true if opt.isUseDropoutLayer if optVars.NumOfLayer ==1 if optVars.isUseBiLSTMLayer opt.layers = [ ... sequenceInputLayer(inputSize) bilstmLayer(optVars.NumOfUnits,'OutputMode','last') dropoutLayer(dropoutVal) fullyConnectedLayer(numResponses) regressionLayer]; else opt.layers = [ ... sequenceInputLayer(inputSize) lstmLayer(optVars.NumOfUnits,'OutputMode','last') dropoutLayer(dropoutVal) fullyConnectedLayer(numResponses) regressionLayer]; end elseif optVars.NumOfLayer==2 if optVars.isUseBiLSTMLayer opt.layers = [ ... sequenceInputLayer(inputSize) bilstmLayer(optVars.NumOfUnits,'OutputMode','sequence') dropoutLayer(dropoutVal) bilstmLayer(optVars.NumOfUnits,'OutputMode','last') dropoutLayer(dropoutVal) fullyConnectedLayer(numResponses) regressionLayer]; else opt.layers = [ ... sequenceInputLayer(inputSize) lstmLayer(optVars.NumOfUnits,'OutputMode','sequence') dropoutLayer(dropoutVal) lstmLayer(optVars.NumOfUnits,'OutputMode','last') dropoutLayer(dropoutVal) fullyConnectedLayer(numResponses) regressionLayer]; end elseif optVars.NumOfLayer ==3 if optVars.isUseBiLSTMLayer opt.layers = [ ... sequenceInputLayer(inputSize) bilstmLayer(optVars.NumOfUnits,'OutputMode','sequence') dropoutLayer(dropoutVal) bilstmLayer(optVars.NumOfUnits,'OutputMode','sequence') dropoutLayer(dropoutVal) bilstmLayer(optVars.NumOfUnits,'OutputMode','last') dropoutLayer(dropoutVal) fullyConnectedLayer(numResponses) regressionLayer]; else opt.layers = [ ... sequenceInputLayer(inputSize) bilstmLayer(optVars.NumOfUnits,'OutputMode','sequence') dropoutLayer(dropoutVal) bilstmLayer(optVars.NumOfUnits,'OutputMode','sequence') dropoutLayer(dropoutVal) bilstmLayer(optVars.NumOfUnits,'OutputMode','last') dropoutLayer(dropoutVal) fullyConnectedLayer(numResponses) regressionLayer]; end elseif optVars.NumOfLayer==4 if optVars.isUseBiLSTMLayer opt.layers = [ ... sequenceInputLayer(inputSize) bilstmLayer(optVars.NumOfUnits,'OutputMode','sequence') dropoutLayer(dropoutVal) bilstmLayer(optVars.NumOfUnits,'OutputMode','sequence') dropoutLayer(dropoutVa