The Altair Community is migrating to a new platform to provide a better experience for you. In preparation for the migration, the Altair Community is on read-only mode from October 28 - November 6, 2024. Technical support via cases will continue to work as is. For any urgent requests from Students/Faculty members, please submit the form linked here
Massive Prediction Problems
der_maik81
Member Posts: 2 
Contributor I
Hey Folks,
I 've gotta big problem. Perhabs someone could give me an advice. Well, I've got a binom. classification problem; 18.000 Examples, 7% of them got the label false and all the others true. I have to get a high recall on the "false" ones.
The decision tree returns only one leaf which is labeled "true". The LibSVm returns just the same (no tree of course). The confusion matrix returns 100% Recall on true and nothing (0%) on false.
That is not what I wanted :-(
I used an xval with libsvm or decision tree (gain ratio) and the performance-measurement "BinomClassificationPerformance"
If you need more information about my experimental setup, i will post it as soon as possible.
Best wishes and lots of thanks,
Maik
I 've gotta big problem. Perhabs someone could give me an advice. Well, I've got a binom. classification problem; 18.000 Examples, 7% of them got the label false and all the others true. I have to get a high recall on the "false" ones.
The decision tree returns only one leaf which is labeled "true". The LibSVm returns just the same (no tree of course). The confusion matrix returns 100% Recall on true and nothing (0%) on false.
That is not what I wanted :-(
I used an xval with libsvm or decision tree (gain ratio) and the performance-measurement "BinomClassificationPerformance"
If you need more information about my experimental setup, i will post it as soon as possible.
Best wishes and lots of thanks,
Maik
0
Answers
did you already try to "tune" your parameters (i.e. prevent radical pruning when learning the decision tree, etc.)? You can also optimize your parameters automatically. That would be my first try to reach a better performance ...
Regards,
Tobias