Prev: Rectangle fitting
Next: How to add legend to an area plot with line plot overlay
From: Kuo-Hsien on 15 May 2010 19:00
Hi all,

I will like to consult your suggestions about your method/suggestions on evaluation of model performance. A model has been modified based a new modeling scheme.

The output from the new modified model is named: output_mod.
The output from the original model is named: output_org.
There is one observation data set is named: obs.

How to evaluate the improvement of the new modified model comparing with the original model? Or, how to highlight the importance of new modeling scheme?

Any suggestions is more than welcome. Thank you.
From: us on 15 May 2010 19:09
"Kuo-Hsien" <mchangks(a)hotmail.com> wrote in message <hsn929$526$1(a)fred.mathworks.com>...
> Hi all,
>
> I will like to consult your suggestions about your method/suggestions on evaluation of model performance. A model has been modified based a new modeling scheme.
>
> The output from the new modified model is named: output_mod.
> The output from the original model is named: output_org.
> There is one observation data set is named: obs.
>
> How to evaluate the improvement of the new modified model comparing with the original model? Or, how to highlight the importance of new modeling scheme?
>
> Any suggestions is more than welcome. Thank you.

where's your problem re ML(?)...

http://www.mathworks.com/products/matlab/

us
From: Kuo-Hsien on 15 May 2010 20:00
> where's your problem re ML(?)...
>
> http://www.mathworks.com/products/matlab/
>
> us

My main questions is: are there any m code to conduct the difference evaluation.
(I need to prove my model is better than the other model.)


From: ImageAnalyst on 15 May 2010 21:23
Maybe generate an ROC curve:
http://en.wikipedia.org/wiki/Receiver_operating_characteristic
which is basically a plot of true positive rate vs. false positive
rate. But it really depends on what kind of model you're talking
about. For example, ROC curves are popular in the medical community
while in compression/decompression, PSNR is popular.
 | 
Pages: 1
Prev: Rectangle fitting
Next: How to add legend to an area plot with line plot overlay