跪求MATLAB 神经网络预测 已贴程序,求解

是 否 +2 论坛币

k人 参与回答

经管之家送您一份

应届毕业生专属福利!

求职就业群

赵安豆老师微信：zhaoandou666

经管之家联合CDA

送您一个全额奖学金名额~ !

立即领取

感谢您参与论坛问题回答

经管之家送您两个论坛币！

+2 论坛币

请问在时间序列数据上怎样预测呢? (如图)

已知x1,x2…为输入变量, x3为输出变量

训练和测试的比例为80/20

当训练完成后要做预测(比如预测之后10个数值), 请问matlab怎样做到?

附上程序, 差最后预测步骤, 求解

1. inputSeries = tonndata(time_in,true,false);
   
2. targetSeries = tonndata(time_out,true,false);
   
3. 
   
4. % Create a Time Delay Network
   
5. inputDelays = 0:1;
   
6. hiddenLayerSize = 10;
   
7. net = timedelaynet(inputDelays,hiddenLayerSize);
   
8. 
   
9. 
   
10. net.inputs{1}.processFcns = {'removeconstantrows','mapminmax'};
    
11. net.outputs{2}.processFcns = {'removeconstantrows','mapminmax'};
    
12. 
    
13. [inputs,inputStates,layerStates,targets] = preparets(net,inputSeries,targetSeries);
    
14. 
    
15. 
    
16. net.divideFcn = 'divideblock';  % Divide data randomly
    
17. net.divideMode = 'time';  % Divide up every value
    
18. net.divideParam.trainRatio = 80/100;
    
19. net.divideParam.valRatio = 20/100;
    
20. net.divideParam.testRatio = 0/100;
    
21. 
    
22. 
    
23. net.trainFcn = 'trainlm';  % Levenberg-Marquardt
    
24. net.performFcn = 'mse';  % Mean squared error
    
25. 
    
26. 
    
27. % Choose Plot Functions
    
28. % For a list of all plot functions type: help nnplot
    
29. net.plotFcns = {'plotperform','plottrainstate','plotresponse', ...
    
30.   'ploterrcorr', 'plotinerrcorr'};
    
31. 
    
32. 
    
33. % Train the Network
    
34. [net,tr] = train(net,inputs,targets,inputStates,layerStates);
    
35. 
    
36. % Test the Network
    
37. outputs = net(inputs,inputStates,layerStates);
    
38. errors = gsubtract(targets,outputs);
    
39. performance = perform(net,targets,outputs)
    
40. 
    
41. % Recalculate Training, Validation and Test Performance
    
42. trainTargets = gmultiply(targets,tr.trainMask);
    
43. valTargets = gmultiply(targets,tr.valMask);
    
44. testTargets = gmultiply(targets,tr.testMask);
    
45. trainPerformance = perform(net,trainTargets,outputs)
    
46. valPerformance = perform(net,valTargets,outputs)
    
47. testPerformance = perform(net,testTargets,outputs)
    
48. 
    
49. % View the Network
    
50. view(net)
    
51. 
    
52. % Plots
    
53. % Uncomment these lines to enable various plots.
    
54. figure, plotperform(tr)
    
55. figure, plottrainstate(tr)
    
56. figure, plotresponse(targets,outputs)
    
57. figure, ploterrcorr(errors)
    
58. figure, plotinerrcorr(inputs,errors)

复制代码

扫码加我 拉你入群

请注明：姓名-公司-职位

以便审核进群资格，未注明则拒绝

分享0 收藏2 回帖

关键词：MATLAB matla atlab 神经网络 Mat 程序 网络

上网找到了以下答案, 请问各位有没有明白以下是甚么意思, 看不懂...

动态时间序列神经网络的预测分成两种情况:

1：循环预测

这种预测非常符合我们的思考方式，毕竟网络的输出是依赖于当前和（或）过去的输入输出，那么同样，y(t+2)是依赖于y(t+1)的，所以循环预测是大家都能接受的方式！

for k=1:2,
  Ai1{1,k}=zeros(10,1);
  Ai1{2,k}=y1{k};
end
for k=1:2,
  Pi1{1,k}=u1{k};
end
yp1 = sim(narx_net_closed,p1,Pi1,Ai1);


2：一次性预测

这种预测方式是Matlab所独特的,因为matlab提供了相应的函数，帮你把数据进行了shift,

[p1,Pi1,Ai1,t1] = preparets(narx_net_closed,u1,{},y1);
yp1 = narx_net_closed(p1,Pi1,Ai1);
plot([cell2mat(yp1)' cell2mat(t1)'])

请问解决了吗，我也遇到这个问题了。