accuracy comparison

[MOHAMMED KAMAL]

Some parts of this message have been removed. Learn more about Nabble's security policy.

 

Eng/Mohammed Kamal

Dear all

i want to compare different data set in terms of accuracy (over all) one number

can i take the percentage of correctly classified only as over all accuracy??

what's the meaning of all these parameters

TP Rate   FP Rate   Precision   Recall  F-Measure   ROC Area

this is example to data set running on j48

thanks in advance

m.kamal

=== Evaluation on test split ===
=== Summary ===

Correctly Classified Instances          47               92.1569 %
Incorrectly Classified Instances         4                7.8431 %
Kappa statistic                          0.8818
Mean absolute error                      0.1603
Root mean squared error                  0.2283
Relative absolute error                 35.987  %
Root relative squared error             48.2891 %
Total Number of Instances               51    

=== Detailed Accuracy By Class ===

TP Rate   FP Rate   Precision   Recall  F-Measure   ROC Area  Class
  1         0          1         1         1          1        Iris-setosa
  0.895     0.063      0.895     0.895     0.895      0.954    Iris-versicolor
  0.882     0.059      0.882     0.882     0.882      0.972    Iris-virginica

=== Confusion Matrix ===

  a  b  c   <-- classified as
 15  0  0 |  a = Iris-setosa
  0 17  2 |  b = Iris-versicolor
  0  2 15 |  c = Iris-virginica

_______________________________________________
Wekalist mailing list
Wekalist@...
https://list.scms.waikato.ac.nz/mailman/listinfo/wekalist

[Peter Reutemann]

> i want to compare different data set in terms of accuracy (over all) one
> number

Use the Experimenter to generate statistically sound results. It makes
it also easier comparing several datasets and/or classifiers.

> can i take the percentage of correctly classified only as over all
> accuracy??

What other accuracy do you have in mind?

> what's the meaning of all these parameters
>
> TP Rate   FP Rate   Precision   Recall  F-Measure   ROC Area

Either read the "Weka Primer" (see FAQ "How do I use Weka from command
line?" for link to article) or the "Data Mining" book by Witten&Frank.

Cheers, Peter
--
Peter Reutemann, Dept. of Computer Science, University of Waikato, NZ
http://www.cs.waikato.ac.nz/~fracpete/ Ph. [phone]

_______________________________________________
Wekalist mailing list
Wekalist@...
https://list.scms.waikato.ac.nz/mailman/listinfo/wekalist