数据：

训练集     tr_117.txt

6.9888 0.0863 4.43E-09 5.4921 3.521 9.056

7.7013 0.0597 7.98E-08 5.4825 2.944 8.886

8.0371 0.4026 2.31E-08 5.2552 3.093 8.886

7.215 0.0864 5.33E-09 5.4841 3.642 8.745

8.7731 0.4815 1.90E-08 5.2548 2.806 8.62

6.5474 0.00E+00 2.30E-10 5.4039 3.421 8.553

8.2223 0.4642 1.49E-08 5.2558 2.778 8.523

6.7448 0.0597 1.03E-08 5.4831 3.222 8.456

6.9867 0.0299 9.88E-09 5.4921 3.32 8.444

6.865 0.0291 9.32E-09 5.3938 3.233 8.444

6.1048 0.0962 1.91E-08 5.2602 3.073 8.444

7.4428 0.1208 9.05E-09 5.4047 3.181 8.398

8.0326 0.4818 8.04E-08 5.2549 3.131 8.347

8.7481 0.4091 1.04E-11 5.257 5.627 8.328

7.1302 0.1071 4.03E-10 5.2563 3.398 8.328

7.2281 0.0827 1.04E-10 5.2584 3.4 8.31

7.9036 0.3771 7.34E-08 5.2554 3.126 8.301

6.2992 0.00E+00 2.18E-10 5.4916 3.654 8.284

8.4513 0.213 4.25E-06 5.2466 3.612 8.276

8.3629 0.5816 1.03E-06 5.2556 3.817 8.252

7.9059 0.299 2.34E-09 5.3941 4.083 8.229

11.9254 2.2228 1.11E-08 5.2593 3.169 8.222

7.7553 0.3858 1.60E-07 5.2552 4.376 8.215

8.3876 0.4459 2.38E-07 5.2542 3.938 8.201

7.6266 0.2015 3.25E-09 5.4871 3.517 8.167

6.949 0.1739 2.38E-08 5.2613 3.07 8.161

8.4303 0.5149 2.52E-08 5.2563 4.302 8.155

12.1249 2.6044 1.26E-07 5.2593 3.79 8.131

7.85 0.32 6.61E-08 5.2549 2.976 8.125

6.8796 0.00E+00 9.82E-09 5.2589 4.15 8.119

8.5383 0.4486 1.59E-07 5.254 3.579 8.108

6.7715 0.1376 1.41E-07 5.2584 4.364 8.086

7.2491 0.2087 1.33E-06 5.2634 3.958 8.076

7.2417 0.3946 3.69E-07 5.264 4.864 8.06

7.1209 0.0932 6.44E-08 5.4693 3.029 8.051

8.2542 0.4007 6.10E-08 5.2548 2.718 8.046

7.7887 0.302 2.74E-07 5.2562 3.374 8.036

7.009 0.1126 5.07E-10 5.3913 3.757 8.032

7.1002 0.0943 3.44E-07 5.4035 2.951 8

7.1016 0.0594 1.60E-08 5.2597 3.545 7.979

10.8391 2.1492 2.30E-10 5.2604 2.858 7.959

6.7052 0.1001 2.15E-11 5.2543 2.887 7.959

7.7881 0.2736 1.04E-11 5.2555 3.769 7.959

7.2796 0.1313 2.14E-11 5.2559 2.567 7.921

8.3152 0.5908 9.36E-08 5.2584 3.16 7.921

9.5298 0.4501 5.69E-09 5.2552 2.805 7.921

7.3635 0.2614 2.82E-06 5.2596 3.385 7.886

6.7722 0.0686 1.03E-07 5.2587 2.044 7.886

8.3582 0.5679 8.26E-07 5.263 3.124 7.854

14.1391 2.2936 8.87E-10 5.2548 3.542 7.854

7.3737 0.0515 1.90E-09 5.2591 4.211 7.854

7.3392 0.1457 3.64E-08 5.2579 2.673 7.796

7.2194 0.1727 1.69E-07 5.2591 6.346 7.796

7.6673 0.37 9.95E-07 5.2938 3.828 7.796

6.652 0.0588 7.74E-09 5.4044 3.006 7.77

6.7773 0.00E+00 1.56E-10 5.2602 3.631 7.721

7.4303 0.2922 8.30E-11 5.2548 4.12 7.721

8.2456 0.396 4.62E-09 5.255 3.533 7.699

8.508 0.3808 3.17E-08 5.2579 2.098 7.658

8.0046 0.3101 4.51E-09 5.2689 3.167 7.658

8.1636 0.6547 3.08E-07 5.2702 3.636 7.638

6.3118 0.5725 3.43E-07 5.3226 3.201 7.638

12.0937 2.4597 1.09E-07 5.2567 2.512 7.62

7.1543 0.2175 7.21E-08 5.2574 3.211 7.62

8.3759 0.6736 7.50E-07 5.2554 3.121 7.569

6.8902 0.0614 6.57E-09 5.2582 3.174 7.509

8.9787 0.4892 1.37E-09 5.2581 1.428 7.509

6.9843 0.0501 2.99E-09 5.2614 3.613 7.481

6.904 0.803 2.28E-07 5.2545 5.195 7.469

6.7145 0.2409 2.26E-08 5.332 3.002 7.469

7.203 0.1296 4.04E-09 5.2587 3.476 7.456

6.7816 0.0603 1.35E-07 5.2595 3.632 7.456

5.8505 0.2795 1.46E-09 5.2537 5.309 7.42

8.0276 0.337 1.16E-08 5.2548 2.731 7.42

12.5221 2.7824 2.37E-09 5.2593 3.941 7.409

7.959 0.4943 1.27E-07 5.2544 3.015 7.387

7.5567 0.2071 6.53E-08 5.2582 3.182 7.377

7.1054 0.0841 1.96E-09 5.4919 3.162 7.337

6.9293 0.1745 0.00E+00 5.2546 4.194 7.301

6.8744 0.0605 3.75E-07 5.2586 3.046 7.292

7.5828 0.2877 4.29E-07 5.2557 3.836 7.26

8.9075 0.5196 1.06E-11 5.2576 1.129 7.244

7.3195 0.245 1.30E-05 5.2513 3.795 7.187

7.1177 0.1667 2.08E-11 5.2587 3.506 7.167

6.529 0.4016 4.90E-09 5.2553 3.324 7.155

11.3517 1.814 5.26E-09 5.2585 3.789 7.149

7.1837 0.1538 1.22E-09 5.2599 3.547 7.137

5.7644 0.1524 6.15E-07 5.2332 3.255 7.114

8.5929 0.7553 7.60E-06 5.2612 5.504 7.086

7.0258 0.3861 2.22E-07 5.256 2.147 7.076

6.9488 2.3122 2.53E-06 5.257 4.877 7.046

7.0021 0.368 1.80E-07 5.2562 2.665 7.032

11.8486 2.2792 2.69E-09 5.2591 3.982 6.921

5.7919 0.2968 1.67E-07 5.2587 4.074 6.921

9.2429 1.248 2.45E-07 5.2584 2.267 6.854

1.5664 0.474 1.14E-08 5.2046 3.26 6.796

8.4639 0.4497 2.62E-08 5.2549 2.16 6.721

2.1186 0.7712 7.73E-08 5.2212 5.631 6.721

8.2095 0.4876 8.86E-09 5.2562 2.623 6.699

1.6618 0.4658 1.90E-08 5.2104 4.46 6.585

6.1497 0.2922 7.91E-09 5.255 3.399 6.444

1.5946 0.4311 3.61E-07 5.239 3.339 6.398

6.4923 2.263 1.71E-06 5.2638 2.475 6.357

1.6025 0.4623 1.83E-06 5.1867 2.951 6.328

2.3521 0.7992 3.69E-07 5.3018 3.664 6.131

1.6074 0.4886 4.22E-07 5.2389 3.241 6.046

1.3386 0.3635 4.62E-06 5.2213 3.483 5.854

7.5329 0.3912 2.45E-05 5.2594 2.881 5.854

6.6985 1.6245 2.31E-07 5.2582 2.655 5.77

6.6985 1.8748 4.03E-07 5.2565 3.628 5.77

1.5056 0.4347 1.80E-07 5.2374 3.21 5.699

1.0759 0.2453 3.02E-07 5.2333 3.062 5.658

1.2281 0.1314 1.07E-06 5.2196 3.707 5.62

1.4192 0.4187 3.78E-06 5.2314 2.787 5.553

1.3169 0.3857 4.69E-06 5.2275 3.232 5.409

1.5337 0.4635 2.93E-06 5.2274 3.239 5.337

1.4303 0.3541 3.66E-07 5.225 2.237 5.076

测试集   ts_29.txt6.8372 0.00E+00 2.11E-09 5.255 3.252 8.658

8.6627 0.3639 5.84E-09 5.2546 2.834 8.456

8.4728 0.476 1.75E-08 5.2555 4.059 8.347

6.8704 0.1493 5.20E-08 5.2595 5.01 8.310

6.9764 0.1108 1.65E-09 5.4891 3.607 8.244

6.3093 0.00E+00 4.18E-11 5.2594 3.593 8.194

8.4666 0.4259 5.06E-09 5.255 3.581 8.131

7.6496 0.3761 5.56E-07 5.2584 3.135 8.081

5.9957 0.107 3.12E-07 5.2602 3.398 8.041

8.7936 0.611 1.50E-07 5.2813 3.877 7.975

7.1682 0.0768 2.15E-11 5.256 2.763 7.921

6.986 0.0268 6.09E-08 5.2599 4.661 7.886

8.7514 0.4551 6.88E-08 5.2581 2.179 7.796

7.2359 0.0907 2.84E-10 5.2592 3.269 7.721

8.2624 0.3756 3.20E-10 5.2791 4.382 7.638

6.5841 0.00E+00 2.13E-07 5.2551 3.449 7.585

6.6638 0.00E+00 3.88E-09 5.2598 3.114 7.469

7.9214 0.5088 4.05E-07 5.255 3.107 7.444

7.3323 0.1151 4.03E-09 5.259 3.113 7.377

4.6289 0.1114 4.78E-10 5.2571 3.641 7.292

8.2305 0.489 1.01E-06 5.2585 3.827 7.155

8.0811 1.2552 2.72E-07 5.2719 3.878 7.097

7.7847 0.4205 1.08E-06 5.2917 4.864 7.000

7.3586 0.0676 7.92E-09 5.2587 3.432 6.721

6.8824 1.6838 5.70E-07 5.2631 2.556 6.569

6.8181 2.4849 1.42E-07 5.258 4.613 6.161

1.426 0.438 5.88E-06 5.2285 2.93 5.796

8.2384 0.4585 4.24E-07 5.2582 3.513 5.658

1.1136 0.422 2.91E-08 5.2093 3.576 5.328

代码如下：

% 清空环境变量clc;

clear all

close all

nntwarn off;

% 录入输入数据load data;

load tr_117.txt;

load ts_29.txt;

tr=tr_117;

ts=ts_29;

[rnum_tr,cnum_tr]=size(tr);

[rnum_ts,cnum_ts]=size(ts);

% 载入数据并将数据分成训练和预测两类

p_train=tr(:,1

cnum_tr-1));

t_train=tr(:,cnum_tr);

p_test=ts(:,1

cnum_ts-1));

t_test=ts(:,cnum_ts);

% 将各个矩阵转置以便适应网络结构

p_train=p_train';

t_train=t_train';

p_test=p_test';

t_test=t_test';

% mapminmax为matlab自带的映射函数

[pn,ps] = mapminmax(p_train);

% 将映射函数的范围参数分别置为1和2

ps.ymin = 0;

ps.ymax = 1;

% 对p_train进行归一化

[pn_tr,ps] = mapminmax(p_train,ps);

% 对pn进行转置,以符合libsvm工具箱的数据格式要求

pn_tr = pn_tr';

% mapminmax为matlab自带的映射函数

[tn_tr,ts] = mapminmax(t_train);

% 将映射函数的范围参数分别置为0和1

ts.ymin = 0;

ts.ymax = 1;

% 对t_train进行归一化

[tn_tr,ts] = mapminmax(t_train,ts);

% 对TSX进行转置,以符合libsvm工具箱的数据格式要求

tn_tr = tn_tr';

% 选择回归预测分析最佳的SVM参数c&g

% 首先进行粗略选择:

% c 的变化范围是 2^(-5),2^(-4),...,2^(10)

% g 的变化范围是 2^(-5),2^(-4),...,2^(5)

[bestmse,bestc,bestg] = SVMcgForRegress(tn_tr,pn_tr,-5,10,-5,5,3,1,1,0.0005);

% 打印粗略选择结果

disp('打印粗略选择结果');

str = sprintf( 'Best Cross Validation MSE = %g Best c = %g Best g = %g',bestmse,bestc,bestg);

disp(str);

% 根据粗略选择的结果图再进行精细选择:

% c 的变化范围是 2^(0),2^(0.3),...,2^(10)

% g 的变化范围是 2^(-2),2^(-1.7),...,2^(3)

[bestmse,bestc,bestg] = SVMcgForRegress(tn_tr,pn_tr,0,10,-2,3,3,0.3,0.3,0.0002);

% 打印精细选择结果

disp('打印精细选择结果');

str = sprintf( 'Best Cross Validation MSE = %g Best c = %g Best g = %g',bestmse,bestc,bestg);

disp(str);

%利用回归预测分析最佳的参数进行SVM网络训练

cmd = ['-c ', num2str(bestc), ' -g ', num2str(bestg) , ' -s 3 -p 0.01'];

model = svmtrain(tn_tr,pn_tr,cmd);

% model = svmtrain(pn,tn,'-s 3 -c 1 -g 2 -p 0.01');

% SVM网络训练集回归预测

[predict,mse] = svmpredict(tn_tr,pn_tr,model);

predict = mapminmax('reverse',predict,ts);

% 打印回归结果

str = sprintf( '均方误差 MSE = %g 相关系数 R = %g%%',mse(2),mse(3)*100);

disp(str);

% 结果分析

figure;

hold on;

plot(t_train,'LineWidth',2);

plot(predict,'r','LineWidth',2);

legend('原始数据','回归预测数据','FontSize',12);

hold off;

grid on;

snapnow;

结果不好，  mse=0.0115523              r2=0.752657

线性的得到的结果和SVM差不多，这个和我以前用R的svm相差很大，一般来讲SVM的结果要好于线性的，r2=0.7318    交叉验证r2=0.7020  ，没有计算测试集的。

还有就是，精细调整 c  g 的范围时候，如何才能知道在哪个范围会更好呢？

还有那个epsilon的加进去的帖子得到的结果都很好，按道理说我的数据也应该不错啊，版主，兄弟们帮忙看看，问题出在哪里？ 谢谢。

[本帖最后由 李洋 于 2009-12-18 11:04 编辑]