glfkuan
2013-03-18, 10:50
为什么我建立的神经网络模型训练后精度不高？请高手指点~~~
数据及matlab程序：
%输入训练数据
>> X=[700 8.8 100.0 214.9 161.6 2 1 2 1 3 4 2
220 2.2 76.3 108.3 100.1 2 1 2 1 3 3 1
80 0.6 50.9 35.3 55.4 3 1 2 1 3 2 2
500 5.8 100.0 98.2 161.6 2 1 3 3 3 3 2
800 5.0 100.0 81.7 161.6 2 2 4 1 3 3 2
350 2.2 91.6 55.8 138.6 3 1 3 1 3 2 3
60 0.3 45.2 28.2 46.8 1 1 1 3 3 3 1
62 0.4 45.8 50.6 48.2 1 1 1 4 4 2 2
110 0.9 57.9 66.6 66.4 3 1 3 1 3 3 2
45 0.4 40.1 33.3 40.1 1 1 2 3 3 3 1
80 1.0 50.9 32.6 55.4 1 1 2 2 2 2 3
60 1.7 45.2 52.7 46.8 1 1 3 3 3 3 1
37 0.5 36.9 70.6 36.4 1 1 3 3 3 4 2
60 1.0 45.2 49.2 46.8 2 1 3 3 4 3 2
38 0.8 37.3 21.9 36.4 1 1 2 3 3 2 1
55 1.0 43.6 36.9 45.4 1 1 2 3 3 1 1
45 0.8 40.1 24.0 40.1 1 2 1 3 3 2 1
100 2.8 55.7 42.9 63.1 3 1 2 3 3 1 1
130 1.2 61.9 75.5 73.1 2 1 3 1 3 2 3
150 4.6 65.6 58.2 80.3 1 1 2 2 3 2 1
200 4.7 73.5 70.4 93.9 2 1 3 2 3 1 1
200 5.3 73.5 137.2 93.9 3 1 2 1 2 4 1
200 2.2 73.5 90.9 93.9 1 2 2 1 2 3 1
700 8.8 100.0 106.0 161.6 2 1 2 1 3 4 1
220 2.2 76.3 54.8 100.1 2 1 2 1 3 2 1
80 0.6 50.9 52.1 55.4 3 1 2 1 3 3 1
500 5.8 100.0 171.0 161.6 2 1 3 3 3 4 3
800 5.0 100.0 194.2 161.6 2 2 4 1 3 3 2
350 2.2 91.6 120.9 138.6 3 1 3 1 3 3 2
60 0.3 45.2 50.0 46.8 1 1 1 3 3 2 1
62 0.4 45.8 28.4 48.2 1 1 1 4 4 2 2
110 0.9 57.9 38.3 66.4 3 1 3 1 3 2 1
45 0.4 40.1 43.0 40.1 1 1 2 3 3 2 1
80 1.0 50.9 57.5 55.4 1 1 2 2 2 3 1
60 1.7 45.2 30.6 46.8 1 1 3 3 3 2 1
37 0.5 36.9 21.9 36.4 1 1 3 3 3 2 1
60 1.0 45.2 27.9 46.8 2 1 3 3 4 2 1
38 0.8 37.3 39.3 36.4 1 1 2 3 3 2 1
55 1.0 43.6 47.4 45.4 1 1 2 3 3 2 1
45 0.8 40.1 42.9 40.1 1 2 1 3 3 2 1];
>> X1=X';
>> V85=[73.7
73.5
53.2
66.7
85.7
72.0
55.8
52.0
72.0
49.4
58.4
57.3
48.0
57.1
50.1
48.3
52.9
57.1
67.9
73.0
71.1
68.0
70.8
65.6
67.9
58.1
90.0
75.0
85.7
50.2
51.0
72.0
52.2
61.0
57.0
50.0
42.9
52.9
46.2
46.2];
>> V852=V85';
%建立神经网络
>> net=newff(minmax(X1),[12,13,1],{'logsig','logsig','purelin'},'trainlm');
net.trainParam.epochs=200;
net.trainparam.goal=0.001;
net=train(net,X1,V852);
求指点！多谢！！！
数据及matlab程序：
%输入训练数据
>> X=[700 8.8 100.0 214.9 161.6 2 1 2 1 3 4 2
220 2.2 76.3 108.3 100.1 2 1 2 1 3 3 1
80 0.6 50.9 35.3 55.4 3 1 2 1 3 2 2
500 5.8 100.0 98.2 161.6 2 1 3 3 3 3 2
800 5.0 100.0 81.7 161.6 2 2 4 1 3 3 2
350 2.2 91.6 55.8 138.6 3 1 3 1 3 2 3
60 0.3 45.2 28.2 46.8 1 1 1 3 3 3 1
62 0.4 45.8 50.6 48.2 1 1 1 4 4 2 2
110 0.9 57.9 66.6 66.4 3 1 3 1 3 3 2
45 0.4 40.1 33.3 40.1 1 1 2 3 3 3 1
80 1.0 50.9 32.6 55.4 1 1 2 2 2 2 3
60 1.7 45.2 52.7 46.8 1 1 3 3 3 3 1
37 0.5 36.9 70.6 36.4 1 1 3 3 3 4 2
60 1.0 45.2 49.2 46.8 2 1 3 3 4 3 2
38 0.8 37.3 21.9 36.4 1 1 2 3 3 2 1
55 1.0 43.6 36.9 45.4 1 1 2 3 3 1 1
45 0.8 40.1 24.0 40.1 1 2 1 3 3 2 1
100 2.8 55.7 42.9 63.1 3 1 2 3 3 1 1
130 1.2 61.9 75.5 73.1 2 1 3 1 3 2 3
150 4.6 65.6 58.2 80.3 1 1 2 2 3 2 1
200 4.7 73.5 70.4 93.9 2 1 3 2 3 1 1
200 5.3 73.5 137.2 93.9 3 1 2 1 2 4 1
200 2.2 73.5 90.9 93.9 1 2 2 1 2 3 1
700 8.8 100.0 106.0 161.6 2 1 2 1 3 4 1
220 2.2 76.3 54.8 100.1 2 1 2 1 3 2 1
80 0.6 50.9 52.1 55.4 3 1 2 1 3 3 1
500 5.8 100.0 171.0 161.6 2 1 3 3 3 4 3
800 5.0 100.0 194.2 161.6 2 2 4 1 3 3 2
350 2.2 91.6 120.9 138.6 3 1 3 1 3 3 2
60 0.3 45.2 50.0 46.8 1 1 1 3 3 2 1
62 0.4 45.8 28.4 48.2 1 1 1 4 4 2 2
110 0.9 57.9 38.3 66.4 3 1 3 1 3 2 1
45 0.4 40.1 43.0 40.1 1 1 2 3 3 2 1
80 1.0 50.9 57.5 55.4 1 1 2 2 2 3 1
60 1.7 45.2 30.6 46.8 1 1 3 3 3 2 1
37 0.5 36.9 21.9 36.4 1 1 3 3 3 2 1
60 1.0 45.2 27.9 46.8 2 1 3 3 4 2 1
38 0.8 37.3 39.3 36.4 1 1 2 3 3 2 1
55 1.0 43.6 47.4 45.4 1 1 2 3 3 2 1
45 0.8 40.1 42.9 40.1 1 2 1 3 3 2 1];
>> X1=X';
>> V85=[73.7
73.5
53.2
66.7
85.7
72.0
55.8
52.0
72.0
49.4
58.4
57.3
48.0
57.1
50.1
48.3
52.9
57.1
67.9
73.0
71.1
68.0
70.8
65.6
67.9
58.1
90.0
75.0
85.7
50.2
51.0
72.0
52.2
61.0
57.0
50.0
42.9
52.9
46.2
46.2];
>> V852=V85';
%建立神经网络
>> net=newff(minmax(X1),[12,13,1],{'logsig','logsig','purelin'},'trainlm');
net.trainParam.epochs=200;
net.trainparam.goal=0.001;
net=train(net,X1,V852);
求指点！多谢！！！