提问人的追问 2011-08-14 13:08 这些没用的，是最上面的数据，这些是为了举个列子。我试了很多遍还是出不来来
回答人的补充 2011-08-14 13:53
主要是组织数据，里面的实际意义你也没详细讲，我这里就针对数据而言。
% 2007-4-30 8.43 10.58 11.47
% 2007-5-31 10.58 11.47 12.67
% 2007-6-29 11.47 12.67 17.31
% 2007-7-31 12.67 17.31 16.52
% 2007-8-31 17.31 16.52 18.5
% 2007-9-28 16.52 18.5 16.33
% 2007-10-31 18.5 16.33 16.62
% 2007-11-30 16.33 16.62 14.48
% 2007-12-28 16.62 14.48 13.26
% 2008-1-31 14.48 13.26 11.14
% 2008-2-29 13.26 11.14 13.32
% 2008-3-31 11.14 13.32 11.56
% 2008-4-30 13.32 11.56 9.04
% 2008-5-30 11.56 9.04 9.33
% 2008-6-30 9.04 9.33 9.03
% 2008-7-31 9.33 9.03 6.42
% 2008-8-29 9.03 6.42 4.83
% 2008-9-26 6.42 4.83 4.93
% 2008-10-31 4.83 4.93 5.45
% 2008-11-28 4.93 5.45 6.86
% 2008-12-31 5.45 6.86 7.31
% 2009-1-23 6.86 7.31 9.01
% 2009-2-27 7.31 9.01 9.52
% 2009-3-31 9.01 9.52
% 2009-4-30 9.52
% 用BP神经网络进行预测
clc; clear all; close all;
format long;
data = [8.43 10.58 11.47
10.58 11.47 12.67
11.47 12.67 17.31
12.67 17.31 16.52
17.31 16.52 18.5
16.52 18.5 16.33
18.5 16.33 16.62
16.33 16.62 14.48
16.62 14.48 13.26
14.48 13.26 11.14
13.26 11.14 13.32
11.14 13.32 11.56
13.32 11.56 9.04
11.56 9.04 9.33
9.04 9.33 9.03
9.33 9.03 6.42
9.03 6.42 4.83
6.42 4.83 4.93
4.83 4.93 5.45
4.93 5.45 6.86
5.45 6.86 7.31
6.86 7.31 9.01
7.31 9.01 9.52];
TT = data(:, 1:2);
t = data(:, 3);
damax=max(max(max(TT)),max(t));
damin=min(min(min(TT)),min(t));
[a,b]=size(TT);
for i=1:a
pp(i,:)=(TT(i,:)-damin)/(damax-damin);
end
tt=(t-damin)/(damax-damin);
n=length(tt);
threshold=[0,1;0,1];
net=newff(threshold,[3,1],{'tansig','logsig'},'trainlm');
net.trainParam.epochs=1000;
net.trainParam.goal=0.001;
LP.lr=0.1;
pp = pp';
net=train(net,pp(:,1:n),tt');
for i=1:length(pp)
p_test=pp(:,i);
out(i)=sim(net,p_test);
pre(i)=out(i)*(damax-damin)+damin;
end
figure; hold on; box on;
plot3(pp(1, 1:n), pp(2, 1:n), tt, 'b-+', pp(1, :), pp(2, :), out, 'r-');
legend('原曲线', '神经网络模拟曲线');
title('神经网络模拟(变换后)');
figure; hold on; box on;
plot3(TT(1:n, 1), TT(1:n, 2), t, 'b-+', TT(:, 1), TT(:, 2), out, 'r-');
legend('原曲线', '神经网络模拟曲线');
title('神经网络模拟(变换前)');
提问人的追问 2011-08-14 14:15
2007-4-308.43
2007-5-3110.58
2007-6-2911.47
2007-7-3112.67
2007-8-3117.31
2007-9-2816.52
2007-10-3118.5
2007-11-3016.33
2007-12-2816.62
2008-1-3114.48
2008-2-2913.26
2008-3-3111.14
2008-4-3013.32
2008-5-3011.56
2008-6-309.04
2008-7-319.33
2008-8-299.03
2008-9-266.42
2008-10-314.83
2008-11-284.93
2008-12-315.45
2009-1-236.86
2009-2-277.31
2009-3-319.01
2009-4-309.52
我用这一排数据来的。用前天昨天今天来预测明天的数据。
这是股票预测。
回答人的补充 2011-08-14 14:22
刚才的代码中，取前两列数据，也就是昨天、今天的数据，来预测目标数据，也就是明天的数据。
% 2007-4-30 8.43 10.58 11.47
% 2007-5-31 10.58 11.47 12.67
% 2007-6-29 11.47 12.67 17.31
% 2007-7-31 12.67 17.31 16.52
% 2007-8-31 17.31 16.52 18.5
% 2007-9-28 16.52 18.5 16.33
% 2007-10-31 18.5 16.33 16.62
% 2007-11-30 16.33 16.62 14.48
% 2007-12-28 16.62 14.48 13.26
% 2008-1-31 14.48 13.26 11.14
% 2008-2-29 13.26 11.14 13.32
% 2008-3-31 11.14 13.32 11.56
% 2008-4-30 13.32 11.56 9.04
% 2008-5-30 11.56 9.04 9.33
% 2008-6-30 9.04 9.33 9.03
% 2008-7-31 9.33 9.03 6.42
% 2008-8-29 9.03 6.42 4.83
% 2008-9-26 6.42 4.83 4.93
% 2008-10-31 4.83 4.93 5.45
% 2008-11-28 4.93 5.45 6.86
% 2008-12-31 5.45 6.86 7.31
% 2009-1-23 6.86 7.31 9.01
% 2009-2-27 7.31 9.01 9.52
% 2009-3-31 9.01 9.52
% 2009-4-30 9.52
% 用BP神经网络进行预测
clc; clear all; close all;
format long;
data = [8.43 10.58 11.47
10.58 11.47 12.67
11.47 12.67 17.31
12.67 17.31 16.52
17.31 16.52 18.5
16.52 18.5 16.33
18.5 16.33 16.62
16.33 16.62 14.48
16.62 14.48 13.26
14.48 13.26 11.14
13.26 11.14 13.32
11.14 13.32 11.56
13.32 11.56 9.04
11.56 9.04 9.33
9.04 9.33 9.03
9.33 9.03 6.42
9.03 6.42 4.83
6.42 4.83 4.93
4.83 4.93 5.45
4.93 5.45 6.86
5.45 6.86 7.31
6.86 7.31 9.01
7.31 9.01 9.52];
TT = data(:, 1:2);
t = data(:, 3);
damax=max(max(max(TT)),max(t));
damin=min(min(min(TT)),min(t));
[a,b]=size(TT);
for i=1:a
pp(i,:)=(TT(i,:)-damin)/(damax-damin);
end
tt=(t-damin)/(damax-damin);
n=length(tt);
threshold=[0,1;0,1];
net=newff(threshold,[3,1],{'tansig','logsig'},'trainlm');
net.trainParam.epochs=1000;
net.trainParam.goal=0.001;
LP.lr=0.1;
pp = pp';
net=train(net,pp(:,1:n),tt');
for i=1:length(pp)
p_test=pp(:,i);
out(i)=sim(net,p_test);
pre(i)=out(i)*(damax-damin)+damin;
end
p_test1 = [9.01 9.52];
p_test2 =(p_test1-damin)/(damax-damin);
out = sim(net, p_test2');
pre1 = out *(damax-damin)+damin
结果
pre1 =
7.498638394693880
也就是，2009-3-31 9.01 9.52对应的结果。
