Ticket #1029 (closed bug: fixed)
TestLearners widget Random Sampling Error
|Reported by:||dimpwnc||Owned by:||anze|
|Severity:||major||Keywords:||random sampling widget testlearners|
The evaluation results for the SVM applied to any dataset (e.g. heart_disease.tab) provides ridiculously low CA and other evaluation results, (e.g. for heart_disease.tab, 0.0527, while cross-validation returns .8382). This is not a product of a single dataset. I initially found this massive discrepancy between cross-validation and random sampling results on my own datasets, but found it to still occur in the standard provided datasets as well.