Re: ARFF vs CSV
Peter Reutemann <[email protected]>
| Newsgroups | gmane.comp.ai.weka |
|---|---|
| Message-ID | <CAHoQ12L_iScHWGwd41Gi0kHkBErdu53D7mdD98UW3HROWa=C_w@mail.gmail.com> |
> I am getting different results depending on whether I open the training set as .arff file or .csv file. > > Is this normal? Many thanks Since you didn't post any details, we can only guess... There are several reasons that I can think of that can influence results: - different order of labels (ARFF predefines order, CSV could have order of appearance or alphabetical) - categorical attribute consisting of numbers is treated as numeric attribute instead when loaded as CSV - the ARFF file format limits the number of decimals (by default) to 6, which can lead an attribute with small numbers change its distribution These can also affect the class attribute (since it is just another attribute). And if you use an algorithm that can act as classifier or regressor, like RandomForest, then switching the class from categorical to numeric will switch how the algorithm is applied. Cheers, Peter -- 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