How to do Single step ahead prediction of Lorenz chaotic time series using Feed Forward Backpropogation Neural Network.

I want to do one step ahead prediction of lorenz chaotic time series data with 3*20000 data.

how to do that.

It will be also help full to get any example to perform one step ahead prediction of chaotic time series using feed forward neural network with delta learning rule.

thanks in advance.

Answer 1

[Greg Heath]

 d = 1
 x = f(:,1:N-d);
 t = f(:,1+d:N);
 rng(0)
 H =10    % default
 net = fitnet(H);
 net.divideFcn = 'dividetrain';
...

 help fitnet
 doc fitnet

Search using

 greg fitnet Ntrials % to find best of multiple designs

[uday]

You are using same data as input and target because x = f(:,1:N+1-d); t = f(:,d:N); will result in same set of value for X and t. so I think using same data for test and train will not do the desired prediction.

[Greg Heath]

Incorrect.

There is only 1 3-D series. Therefore, the only difference between input and target is a shift in the data. When the input is f(:,1), the predicted output should be f(:,1+d). Therefore, the correct answer is

 x =f(:,1:N-d); t = f(:,1+d:N);

I will make the appropriate correction in my accepted answer.

Hope this helps.

Greg

Answer 2

[Greg Heath]

 help narnet

 doc narnet

For practice

 help nndatasets

Also search the NEWSGROUP and ANSWERS using

 greg narnet

Hope this helps

Thank you for formally accepting my answer

Greg

[uday]

Actually I want to do it using feedforwardnet only not using any time series analysis tools like NAR or NARX.

[Greg Heath]

See my first answer.