Prev: N-gram extraction
Next: lscov vs \
From: david on 25 May 2010 04:34
Hello all,
I get a root mean sraure error RMSE =6.896 from time series forecasting by a neural networks . I want to know if this error is acceptable for such problem or it is too hight and it needs to improve .

Thanks in advance

David
From: Greg Heath on 25 May 2010 07:17
On May 25, 4:34 am, "david " <david.sabine...(a)gmail.com> wrote:
> Hello all,
> I get a root mean sraure error RMSE =6.896 from time series forecasting
> by a neural networks . I want to know if this error is  acceptable for such  
> problem or it is too hight and it needs to improve .

Use the R^2 statistic for validation or test data.
Use the Ra^2 (degree-of-freedom adjusted R^2) statistic
for training data.

Set net.train.Param.goal to obtain Ra^2 = 0.99

See my previous posts in comp.soft-sys.matlab for details.
For example, search in CSSM using

greg heath R2

Hope this helps.

Greg
From: Stephen Vanreusel on 25 May 2010 09:33
Greg Heath <heath(a)alumni.brown.edu> wrote in message <dc9913dc-d184-499d-b78b-78f2c471d19a(a)i31g2000vbt.googlegroups.com>...
> On May 25, 4:34 am, "david " <david.sabine...(a)gmail.com> wrote:
> > Hello all,
> > I get a root mean sraure error RMSE =6.896 from time series forecasting
> > by a neural networks . I want to know if this error is  acceptable for such  
> > problem or it is too hight and it needs to improve .
>
> Use the R^2 statistic for validation or test data.
> Use the Ra^2 (degree-of-freedom adjusted R^2) statistic
> for training data.
>
> Set net.train.Param.goal to obtain Ra^2 = 0.99
>
> See my previous posts in comp.soft-sys.matlab for details.
> For example, search in CSSM using
>
> greg heath R2
>
> Hope this helps.
>
> Greg

Also, which version of MATLAB are you using? As of R2008a, you can view your results on the NNTRAINTOOL plot produced by the training process. The "Plot Regression" pushbutton allows you to see the regression plots for the training, validation and test data.

Also, you can do the same by using the PLOTREGRESSION function on your validation data.

Hope this helps,
Steve
 | 
Pages: 1
Prev: N-gram extraction
Next: lscov vs \