I want to forecast for next period

I read source code of nax neural network on http://in.mathworks.com/help/fuzzy/examples/chaotic-time-series-prediction.html?refresh=true but i dntd find forecasting time series for next period exampl for one month or more or less plz can any one help to get the source for forecasting next period or understanding

thank u

Answer 1

[Greg Heath]

Since you obviously wrote NAX in error, it is not clear whether you meant NAR or NARX. Assuming NAR,

1. Start with the documentation and simple example

 help narnet
 doc narnet

2. Consider the other NARNET examples

 help nndatasets
 doc nndatasets

3. Search the NEWGROUP and ANSWERS for relevant posts

NARNET
NARNET tutorial
NARNET greg
NARNET tutorial greg

Hope this helps.

Greg

[abduall hassn]

Greg sir i am relevant your posts first of all i am sorry of typing mistake means NARX

Greg sir i am relevant your posts but iam getting forecasting next period sir i have network data for 3 years from 2005-2010 i want to forecasting 2011 data this is i want plz if u can explain for me this part thank u sir alot

Answer 2

[Greg Heath]

2005 to 2010 is 5 years, not 3.

 [ I N ] = size(input)% = ?
 [ O N ] = size(target)% = ?

Try divideblock with a (4/6)/(1/6)/(1/6) data division over 6 years.

Sounds straightforward using some of my posted code.

Do not use test data to estimate significant lags, number of hidden nodes, etc.

Hope this helps.

Greg

[abduall hassn]

Thank you very much for your answer Greg sir

and Sir one more help

My data is 1x2551 i have only one attribute

i applied this code for forecasting next 30 days

i want to know this code is corrected and how i can find error by using MAPE NAD MSE

   net = narnet(1:4,10);

net.trainParam.showWindow = false;

 T = tonndata(T,false,false);

 [Xs,Xi,Ai,Ts] = preparets(net,{},{},T);

net = closeloop(train(net,Xs,Ts,Xi,Ai));

 Ypred = nan(30,1);

Ypred = tonndata(Ypred,false,false);

Ypred(1:4) = T(end-3:end);

[xc,xic,aic,tc] = preparets(net,{},{},Ypred);

 Ypred = fromnndata(net(xc,xic,aic),true,false,false);
  end

[Greg Heath]

Apply your code to one of the NARNET example data sets other than the simplenar-dataset set and I will compare results.

 help nndatasets
 doc nndatasets

 chickenpox_dataset- 498 Monthly chickenpox instances dataset 
 ice_dataset       - 219 Global ice volume dataset
 laser_dataset     - 10093 Chaotic far-infrared laser dataset
 oil_dataset       - 180  Monthly oil price dataset.
 river_dataset     - 264  River flow dataset.
 solar_dataset     - 2899 Sunspot activity dataset.

Greg

[Greg Heath]

How often are your 2551 measurements ?

 Daily: 2551/365.25 ~ 7 years

So you want to use 2521 to predict the last 30?

[abduall hassn]

Greg sir yes

it is 7 years

no i want to forecast 2581 means next 30 days

[Greg Heath]

OK. However, how will you know if the answer is correct?

If you don't, I suggest first doing 2521 + 30. Once that works then you can update the net with the last 30 and then predict 2552 t0 2581.

Have you used the autocorrelation function to find the significant lags?

[abduall hassn]

sir i used autocorrelation function for ARIMA model for determine lags

 but, neural network  i didnt used  autocorrelation function

[abduall hassn]

 Bro  Greg ,  i have   work as u said   521 + 30.    It  got  good prediction but the problem

How i can upadate the net with last.

And thanking for ur help.