Re: Confusion Matirx

Depending on your classifier (which you didn't post), the model probably only consists of a single split point. After all, there is only a single variable to build a model on.

Have you tried using discretization (supervised or unsupervised filter) on your real valued attribute? You can wrap your classifier and the filter in the FilteredClassifier meta classifier. Then you don't have to worry about dataset compatibility, you can just use the original data.

Cheers, Peter

On October 16, 2021 6:14:25 PM GMT+13:00, Bob Matthews <[email protected]> wrote:
>=== Summary ===
>
>Correctly Classified Instances         376 62.5624 %
>Incorrectly Classified Instances       225 37.4376 %
>Kappa statistic                          0
>Mean absolute error                      0.4684
>Root mean squared error                  0.484
>Relative absolute error                100      %
>Root relative squared error            100      %
>Total Number of Instances              601
>
>=== Detailed Accuracy By Class ===
>
>                   TP Rate  FP Rate  Precision  Recall F-Measure
>MCC      ROC Area  PRC Area  Class
>                   1.000    1.000    0.626      1.000    0.770 ?
>0.500     0.626     TRUE
>                   0.000    0.000    ?          0.000    ? ? 0.500
>0.374     FALSE
>Weighted Avg.    0.626    0.626    ?          0.626    ? ? 0.500     0.532
>
>=== Confusion Matrix ===
>
>     a   b   <-- classified as
>   376   0 |   a = TRUE
>   225   0 |   b = FALSE
>
>When testing against the test set(30%) all predictions are TRUE !!!
>
>I am using J48 -C 0.25 -M 2
>
>My instances comprise to values:-
>
>(a) a real value (decimal)
>
>(b) a nominal value (TRUE/FALSE)
>
>My question: What does it mean when ALL predictions are "TRUE"
>
>How do I improve so that the predictions are either "TRUE" or "FALSE"
>
>Bob M
>

--
Peter Reutemann
Dept. of Computer Science
University of Waikato, NZ
+64 (7) 858-5174 (office)
+64 (7) 577-5304 (home office)
http://www.cms.waikato.ac.nz/~fracpete/
http://www.data-mining.co.nz/.

_______________________________________________
Wekalist mailing list -- [email protected]
Send posts to [email protected]
To unsubscribe send an email to [email protected]
To subscribe, unsubscribe, etc., visit https://list.waikato.ac.nz/postorius/lists/wekalist.list.waikato.ac.nz
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html