B&Em

%B

clear all;
close all;
clc;
% 导入训练数据-----------------------
P = xlsread('777','sheet1','A2:G221');
%T = xlsread('777','sheet1','H2:H221');
T = xlsread('888','sheet1','C2:C221');
P = P';
T = T';

%设置训练样本集
trainx = P(:,1:200);
trainy = T(1:200);
%设置测试样本集
testx = P(:,201:220);
testy = T(201:220);

% 数据归一化
% 训练集
[Pn_train,inputps] = mapminmax(trainx);
[Tn_train,outputps] = mapminmax(trainy);

% 测试集
Pn_test = mapminmax('apply',testx,inputps);
Tn_test = mapminmax('apply',testy,outputps);
%------------------------------------
%创建神经网络
%net_1 = newff(minmax(P),[12,4],{'tansig','purelin'},'traingdm');

net_1 = newff(Pn_train,Tn_train,[7],{'tansig','purelin'},'traingdm');

%  当前输入层权值和阈值
inputWeights=net_1.IW{1,1};
inputbias=net_1.b{1};
%  当前网络层权值和阈值
layerWeights=net_1.LW{2,1};
layerbias=net_1.b{2};
%设置训练参数
net_1.trainParam.show = 50;  %每间隔100步显示一次训练结果
net_1.trainParam.lr = 0.01;  %学习速率0.05
net_1.trainParam.mc = 0.9;
net_1.trainParam.epochs = 20000; %允许最大训练步数1000步
net_1.trainParam.goal = 1e-3;  %训练目标最小误差0.001

%训练网络
[net_1,tr] = train(net_1,Pn_train,Tn_train);
%%
%使用训练好的网络，自定义输入
%A = sim(net_1,P);
%理想输出与训练输出的结果进行比较
%E = T - A;
%计算误差
%MSE = mse(E)
%%
%第二组验证
A1 = sim(net_1,Pn_test);
%反归一化
yy = mapminmax('reverse',A1,outputps);
E1 = testy' - yy';
MSE1 = mse(E1)

%绘图
hold on;

figure(1);
%subplot(211);

plot(testy','-*');
plot(yy','-.gp');
legend('理论值','实际值');
hold off;

figure(2);

plot(E1','-or');
grid on;
xlabel('E1 = T1 - A1; ');ylabel('误差曲线');

%如果你对该次训练满意，可以保存训练好网络
%save('123.mat','net_1','tr');

<pre name="code" class="cpp">%Em神经网络预测clear all;close all;clc;% 导入训练数据-----------------------P = xlsread('777','sheet1','A2:G221');%T = xlsread('777','sheet1','H2:H221');T = xlsread('888','sheet1','C2:C221');P = P';T = T';%设置训练样本集trainx = P(:,1:200);trainy = T(1:200);%设置测试样本集testx = P(:,201:220);testy = T(201:220);testx1 = P(:,101:120);testy1 = T(101:120);% 数据归一化% 训练集[Pn_train,inputps] = mapminmax(trainx);[Tn_train,outputps] = mapminmax(trainy);% 测试集Pn_test = mapminmax('apply',testx,inputps);Tn_test = mapminmax('apply',testy,outputps);%------------------------------------%创建神经网络net_1 = newelm(Pn_train,Tn_train,[12,4],{'tansig','purelin'},'traingdm');%设置训练参数net_1.trainParam.show = 50;net_1.trainParam.lr = 0.05;net_1.trainParam.mc = 0.9;net_1.trainParam.epochs = 10000;net_1.trainParam.goal = 1e-3;%训练网络net_1 = train(net_1,Pn_train,Tn_train);%测试网络%第1组验证%使用训练好的网络，自定义输入A1 = sim(net_1,Pn_test);%反归一化yy = mapminmax('reverse',A1,outputps);E1 = testy' - yy'; %理想输出与训练输出的结果进行比较%MSE1 = mse(E1) %计算误差%绘图hold on;figure(1);%subplot(211);plot(testy','-*');plot(yy','-.gp');legend('理论值','预测值');hold off;figure(2);plot(E1','-or');grid on;xlabel('E1 = T1 - A1; ');ylabel('误差曲线');