MATLAB使用BP神经网络评估输入指标对结果的影响大小_bp神经网络输入因子对输出的影响比较

p=[0.215703	0.184468	0.669382	0.071882;
0.508849	0.613202	0.010964	0.005636;
0.018413	0.312329	0.015825	1.000000;
0.198495	0.168350	0.012520	0.002022;
0.241937	0	        0.790496	0.0718815;
0.10794	        0.276951	0.003350	0.022384;
0.3269415	0.1682795	0.834838	0.2510915;
0.5	        0.625	        0.811576	0.17779]
 
F=[50.40044478	198.3434254	219.2033856	65.36585321	170.7157683	64.71545211	130.3209169	144.0268874
];
 
t=F    % 此时保证p为一行一个样本，t表示的结果为一行数据
 
[m,n]=size(p);
yy_temp = p;
for i=1:n
p=yy_temp;
pX=p(:,i);
pa=pX*1.1;
p(:,i)=pa;
aa=['p_increase' int2str(i) '=p'];
eval(aa);
end
 
for i=1:n
p=yy_temp;
pX=p(:,i);
pa=pX*0.9;
p(:,i)=pa;
aa=['p_decrease' int2str(i) '=p'];
eval(aa);
end
 
nntwarn off;
p=p';
 
net = newff(minmax(p),[8 1],{'tansig', 'purelin'}, 'traingdm');
net=init(net);
net.trainParam.show = 50;
net.trainParam.lr = 0.05;
net.trainParam.mc = 0.9;
net.trainParam.epochs = 2000;
net = train(net,p,t);
 
for i=1:n
eval(['p_increase', num2str(i), '=transpose(p_increase', num2str(i), ')'])
end
 
for i=1:n
eval(['p_decrease', num2str(i), '=transpose(p_decrease', num2str(i), ')'])
end
 
for i=1:n
eval(['result_in',num2str(i),'=sim(net,','p_increase',num2str(i),')'])
end
 
for i=1:n
eval(['result_de',num2str(i),'=sim(net,','p_decrease',num2str(i),')'])
end
 
for i=1:n
eval(['result_in',num2str(i),'=transpose(result_in',num2str(i),')'])
end
 
for i=1:n
eval(['result_de',num2str(i),'=transpose(result_de',num2str(i),')'])
end
 
for i=1:n
IV=['result_in',num2str(i),'-result_de',num2str(i)];
eval(['MIV_',num2str(i),'=mean(',IV,')'])
end
 
 

结果输出的MIV即为输入指标对结果的影响大小，如果输出MIV全零，注意将p归一化。

具体见[MATLAB神经网络30个案例分析].史峰.扫描版[www.xuexi111.com].pdf中第20个案例。