Name | Modified | Size | Downloads / Week |
---|---|---|---|
Parent folder | |||
tmc2007-500-train.arff | 2014-09-30 | 3.2 MB | |
tmc2007-500-test.arff | 2014-09-30 | 1.0 MB | |
README.txt | 2014-09-30 | 661 Bytes | |
yeast-train.arff | 2014-09-30 | 1.5 MB | |
yeast-test.arff | 2014-09-30 | 918.5 kB | |
scene-train.arff | 2014-09-30 | 3.1 MB | |
scene-test.arff | 2014-09-30 | 3.1 MB | |
music-test.arff | 2014-09-30 | 133.0 kB | |
medical-train.arff | 2014-09-30 | 98.9 kB | |
genbase-train.arff | 2014-09-30 | 1.7 MB | |
genbase-test.arff | 2014-09-30 | 758.7 kB | |
enron-train.arff | 2014-09-30 | 607.8 kB | |
enron-test.arff | 2014-09-30 | 324.2 kB | |
emotions-train.arff | 2014-09-30 | 252.1 kB | |
emotions-test.arff | 2014-09-30 | 132.0 kB | |
Totals: 15 Items | 16.8 MB | 1 |
These datasets are supplied in train-test format; as also in http://mulan.sourceforge.net/datasets-mlc.html The are good for comparing to e.g., Madjarov et al "An extensive experimental comparison of methods for multi-label learning" (among many others). Here the -C flag is set into the header with the number of labels for Meka, so it does not need to be specified at run time. Simply use the flags -t emotions-train.arff -T emotions-test.arff Note that these datasets are in 'Mulan format' (with the classes at the end). Meka will automatically move them to the beginning before classification. This may take a few extra seconds for very large datasets.