基于遗传优化的双BP神经网络金融序列预测算法matlab仿真

1.程序功能描述
基于遗传优化的双BP神经网络金融序列预测算法matlab仿真，采用的双BP神经网络结构如下：

2.测试软件版本以及运行结果展示
MATLAB2022A版本运行

三个算法的误差对比：

3.核心程序

LEN = 10;
%样本的划分
for i = 1:length(C)-LEN
    Price1(:,i) = C(i:i+LEN-1);
    Price2(i)   = C(i+LEN);
end
 
%训练样本
L1 = floor(0.6*length(Price2));
for i = 1:L1
    train_data(:,i) = Price1(:,i); 
    train_aim(i)    = Price2(i);
end
 
%测试样本
L2 = length(Price2) - L1;
for i = 1:L2
    test_data(:,i) = Price1(:,i+L1); 
    test_aim(i)    = Price2(i+L1);
end
 
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%如下的是改进BP网络算法
%定义神经网络的各个层的个数
Num_In     = LEN;
Num_Hidden = 60;
Num_Out    = 1;
%构建BP网络
net        = newff(train_data,train_aim,Num_Hidden);
 
ERR1 = [];
ERR2 = [];
ERR3 = [];
for j = 1:5
    j
    %通过改进遗传算法优化BP参数
    net        = func_newGA(net,Num_In,Num_Hidden,Num_Out,train_data,train_aim);
    %网络训练
    net.trainParam.showWindow = 0;
    net        = train(net,train_data,train_aim);
    outputs    = sim(net,test_data);
    d1         = test_aim*(Maxs-Mins) + Mins;
    d2         = outputs*(Maxs-Mins) + Mins;
    ERR1       = [ERR1,mean(abs(d1-d2)./d2) ];
    ERR2       = [ERR2,mean((abs(d1-d2)./d2).^2) ];
    ERR3   = [ERR3,std((abs(d1-d2)./d2).^2) ];
end
 
figure;
plot(d1,'b');
hold on
plot(d2,'r');
legend('真实股价','预测股价');
xlabel('时间（days）');
ylabel('收盘价格对比');
 
disp('平均误差：');
mean(ERR1)
disp('平方差：');
mean(ERR2)
disp('均方差：');
mean(ERR3)
save r2.mat d1 d2

4.本算法原理
整个BP算法的整体结构如下所示：