机器学习之MATLAB代码--CNN预测 -慈云数据

机器学习之Matlab代码--CNN预测 _LSTM预测（十七）

代码

数据

结果

代码

下列代码按照下列顺序依次：

在这里插入图片描述

1、

clc
clear all
%%
load('Test.mat')
Test(1,:) =[];
YTest = Test.demand;
XTest = Test{:,3:end};
%% LSTM CNN CNN-LSTM 
[LSTM_YPred]=LSTM_Predcit();
[CNN_YPred]=CNN_Predcit();
[CNN_LSTM_YPred]=CNN_LSTM_Predcit();
%% 鐢诲浘姣旇緝
figure
plot(LSTM_YPred,'r')
hold on 
plot(CNN_YPred,'b')
hold on 
plot(CNN_LSTM_YPred,'y');  
hold on
plot(YTest,'k')
legend('LSTM','CNN','CNN-LSTM','实际')

2、

function [LSTM_YPred] = LSTM_Predcit()
%% 数据初始化
load('Train.mat')
Train(1,:) =[];
YTrain = Train.demand;
XTrain = Train{:,3:end};
load('Test.mat')
Test(1,:) =[];
YTest = Test.demand;
XTest = Test{:,3:end};
%% 训练集数据归一化
XTrain = XTrain';
YTrain = YTrain';
[XTrainStandardized,Xps_in] = mapminmax(XTrain,0,1);
[YTrainStandardized,Yps_out] = mapminmax(YTrain,0,1);
%% 创建神经网络
layers = [
    sequenceInputLayer(6,"Name","input")
    lstmLayer(128,"Name","gru")
    dropoutLayer(0.1,"Name","drop")
    fullyConnectedLayer(1,"Name","fc")
    regressionLayer("Name","regressionoutput")];
options = trainingOptions( 'adam',...
           'MaxEpochs',50,...
           'GradientThreshold',1,...
           'InitialLearnRate',0.005,...
           'LearnRateSchedule','piecewise',...
           'LearnRateDropPeriod', 125,...
           'LearnRateDropFactor' ,0.2,...
           'Verbose' , 0,...
           'MiniBatchSize',2,...
           'Plots','training-progress' );
%% 训练网络
net= trainNetwork(XTrainStandardized,YTrainStandardized,layers,options);
%% 测试集归一化
XTest = XTest';
XTestStandardized = mapminmax('apply',XTest,Xps_in);
%% 多步预测
net = predictAndUpdateState(net,XTrainStandardized);
numTimeStepsTest = numel(XTest(1,:));
LSTM_YPred = [];
for i = 1:numTimeStepsTest
    [net, LSTM_YPred(i)] = predictAndUpdateState(net,XTestStandardized(:,i),'ExecutionEnvironment','cpu');
end
LSTM_YPred = mapminmax('reverse',LSTM_YPred,Yps_out);
%% 多步预测
net = predictAndUpdateState(net,XTrainStandardized);
numTimeStepsTest = numel(XTest(1,:));
LSTM_YPred = [];
for i = 1:numTimeStepsTest
    [net, LSTM_YPred(i)] = predictAndUpdateState(net,XTestStandardized(:,i),'ExecutionEnvironment','cpu');
end
LSTM_YPred = mapminmax('reverse',LSTM_YPred,Yps_out);
%% 结果可视化
%绘图
figure 
plot(LSTM_YPred,'g-')
hold on 
plot(Test.demand);    
legend('预测值','实际值')

3、

function [CNN_YPred] = CNN_Predcit()
%% 数据初始化
load('Train.mat')
Train(1,:) =[];
XTrain = (Train{:,3:end})';
YTrain = (Train.demand)';
load('Test.mat')
Test(1,:) =[];
XTest = (Test{:,3:end})';
YTest = (Test.demand)';
M = size(XTrain,2);
N = size(XTest,2);
%% 训练集数据归一化
[X_train,ps_input] = mapminmax(XTrain,0,1);
X_test = mapminmax('apply',XTest,ps_input);
[Y_train,ps_output] = mapminmax(YTrain,0,1);
Y_test = mapminmax('apply',YTest,ps_output);
%% 数据平铺
X_train = double(reshape(X_train,6,1,1,M));
X_test  = double(reshape(X_test,6,1,1,N));
Y_train = double(Y_train)';
Y_test  = double(Y_test)';
%% %% 构建网络结构
layers = [
    imageInputLayer([6 1 1],"Name","imageinput")                 %输入
    convolution2dLayer([3 1],16,"Name","conv","Padding","same")  %卷积
    batchNormalizationLayer("Name","batchnorm")                  %归一
    reluLayer("Name","relu")                                     %激活 
 
    convolution2dLayer([3 1],32,"Name","conv_1","Padding","same")
    batchNormalizationLayer("Name","batchnorm_1")
    reluLayer("Name","relu_1")
    dropoutLayer(0.2,"Name","drop")
    fullyConnectedLayer(1,"Name","fc")                           %全连接输出
    regressionLayer("Name","regressionoutput")];                 %回归
%% 参数设置
options = trainingOptions( 'sgdm',...            %SGDM梯度下降算法
           'MiniBatchSize',30,...                %批大小，每次训练样本个数30
           'MaxEpochs',10,...                   %最大训练次数800
           'InitialLearnRate',0.01,...           %初始学习率0.01
           'LearnRateSchedule','piecewise',...   %学习率下降
           'LearnRateDropFactor' ,0.5,...        %学习率下降因子
           'LearnRateDropPeriod', 400,...        %经过400次训练后 学习率0.01*0.5
           'Shuffle','every-epoch',...           %每次训练打乱数据集
           'Plots','training-progress',...       %画出曲线
           'Verbose',false);
%% 训练模型
net= trainNetwork(X_train,Y_train,layers,options);
%% 模型预测
CNN_YPred = predict(net,X_test);
%% 数据反归一化
CNN_YPred = mapminmax('reverse',CNN_YPred,ps_output);
%% 均方根误差
error2 = sqrt(sum((CNN_YPred'-Y_test).^2)./N);
%% 绘制网路分布图
analyzeNetwork(layers)
%% 绘图
figure
plot(1:N, YTest,'r-',1:N,CNN_YPred, 'b','Linewidth', 1)
legend('真实值','预测值')
xlabel('预测样本')
ylabel('预测结果')
string = {'测试集预测结果对比';['RMSE=' num2str(error2)]};
title(string)
xlim([1,N])
grid
% %%相犬指标计算
% %R2
% R2 = 1 - norm(T_test - Y_Pred')^2 / norm(T_test - mean(T_test))^2;
% disp(['测试集数据的R2为: ', num2str(R2)])
% %MAE
% mae2 = sum( abs(Y_Pred' - T_test )) ./ N ;
% disp( ['测试集数据的MAE为: ', num2str(mae2)])
% %MBE
% mbe2 = sum( Y_Pred' - T_test ) ./ N ;
% disp(['测试集数据的MBE为: ', num2str(mbe2)])
figure
plot(CNN_YPred)
hold on
plot(YTest)
legend('预测值','实际值')

4、

function [CNN_LSTM_YPred] = CNN_LSTM_Predcit()
%% 数据集（列为特征，行为样本数目 6特征1输出)
clc
clear all
%
load('Train.mat')
Train(1,:) =[];
XTrain = Train{:,3:end};
YTrain = Train.demand;
[XTrain_norm,Xopt] = mapminmax(XTrain',0,1);
[YTrain_norm,Yopt] = mapminmax(YTrain',0,1);
% X = X';
k = 24; % 滞后长度
% 转换成2-D image(数据集平铺)
for i = 1:length(YTrain_norm)-k
    Train_XNorm{i} = reshape(XTrain_norm(:,i:i+k-1),6,1,1,k);
    Train_YNorm(i) = YTrain_norm(i+k-1);
end
Train_YNorm= Train_YNorm';
%  
load('Test.mat')
Test(1,:) =[];
YTest = Test.demand;
XTest = Test{:,3:end};
[XTestnorm] = mapminmax('apply', XTest',Xopt);
[YTestnorm] = mapminmax('apply',YTest',Yopt);
XTest = XTest';
for i = 1:length(YTestnorm)-k
    Test_XNorm{i} = reshape(XTestnorm(:,i:i+k-1),6,1,1,k);
    Test_YNorm(i) = YTestnorm(i+k-1);
    Test_Y(i) = YTest(i+k-1);
end
Test_YNorm = Test_YNorm';
%% LSTM 层设置，参数设置
inputSize = size(Train_XNorm{1},1);   %数据输入x的特征维度
outputSize = 1;  %数据输出y的维度  
numhidden_units1=50;
numhidden_units2= 20;
numhidden_units3=100;
opts = trainingOptions('adam', ...
    'MaxEpochs',10, ...
    'GradientThreshold',1,...
    'ExecutionEnvironment','cpu',...
    'InitialLearnRate',0.001, ...
    'LearnRateSchedule','piecewise', ...
    'LearnRateDropPeriod',2, ...   %2个epoch后学习率更新
    'LearnRateDropFactor',0.5, ...
    'Shuffle','once',...  % 时间序列长度
    'SequenceLength',k,...
    'MiniBatchSize',24,...
    'Plots','training-progress',...
    'Verbose',0);
%% CNN-LSTM
layers = [ ...
    
    sequenceInputLayer([inputSize,1,1],'name','input')   %输入层设置
    sequenceFoldingLayer('name','fold')
    convolution2dLayer([2,1],10,'Stride',[1,1],'name','conv1')
    batchNormalizationLayer('name','batchnorm1')
    reluLayer('name','relu1')
    maxPooling2dLayer([1,3],'Stride',1,'Padding','same','name','maxpool')
    sequenceUnfoldingLayer('name','unfold')
    flattenLayer('name','flatten')
    gruLayer(numhidden_units1,'Outputmode','sequence','name','hidden1') 
    dropoutLayer(0.3,'name','dropout_1')
    lstmLayer(numhidden_units2,'Outputmode','last','name','hidden2') 
    dropoutLayer(0.3,'name','drdiopout_2')
    fullyConnectedLayer(outputSize,'name','fullconnect')   % 全连接层设置（影响输出维度）（cell层出来的输出层） %
    tanhLayer('name','softmax')
    regressionLayer('name','output')];
lgraph = layerGraph(layers)
lgraph = connectLayers(lgraph,'fold/miniBatchSize','unfold/miniBatchSize');
%% 网络训练
tic
net = trainNetwork(Train_XNorm,Train_YNorm,lgraph,opts);
%% 绘制网路分布图
analyzeNetwork(layers)
%% 测试
CNN_LSTM_YPred_norm = net.predict(Test_XNorm);
CNN_LSTM_YPred  = mapminmax('reverse',CNN_LSTM_YPred_norm',Yopt);
CNN_LSTM_YPred = CNN_LSTM_YPred';
%%
close all
figure
plot(CNN_LSTM_YPred,'g-')
hold on 
plot(Test_Y);    
legend('预测值','实际值')