Re: Clarification on MultiClassClassifier
Eibe Frank <[email protected]>
| Newsgroups | gmane.comp.ai.weka |
|---|---|
| Message-ID | <[email protected]> |
Yes, the MultiClassClassifier turns a multi-class problem into several binary-class problems. By default, it uses the one-vs-rest method. However, it can also optionally (a) perform one-vs-one (aka pairwise) classification (and, with a further option turned on, pairwise coupling), or (b) use random error-correcting output codes, or (c) use exhaustive error-correcting output codes. In each case, binary classification problems are created. You can look at the classifiers that are generated in the output (assuming the base classifier is a decision tree or something that has a nice printable representation). Cheers, Eibe > On 8/06/2021, at 12:40 AM, [email protected] wrote: > > Hi, > looking at the documentation [https://www.cs.waikato.ac.nz/ml/weka/Witten_et_al_2016_appendix.pdf] it seems that whenever a MultiClassClassifier is used, the multi-class problem is treated as a binary one (I guess comparing one class vs the rest). However, there is no reference on how the problem is handled without toning the base classifier. To be clear: > 1. if I use i.e. a Random Forest or J48 classifiers for multiclass classification, is the problem treated as a binary one by default? > 2. How can I understand if the problem is handled as binary or multiclass? > _______________________________________________ > 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 _______________________________________________ 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