The documentation said use Classify, you will get state of the art machine learning result. But since the Classify function of Mathematica work much like a black box, we do not do manual feature engineering like imputation missing data, feature selection etc.
So I want to know is there any benchmark to show that Mathematica Classify function actually doing as good as or even better than using other machine library like sklearn and do manual feature engineering?
Classifyfunction is, whether there is a benchmark to measure it. Whether it actually do better than manual feature engineering approach like using python. The other question I want to know whether we can know the internal pre-processing mechanism. Whether we can specify the method like to explicit indicate the classify function to use inClassifyfunction – m00nlight Oct 07 '15 at 09:56Classify. Furthermore, in both cases you did not explain what exactly is that you cannot find in the documentation (see e.g. Classify > Options > Method). I think it would be rather vague to compare any reasonable banchmarking results considering the many options you can choose for such a high-level function. Again, this boils down to the internals of the function. – István Zachar Oct 07 '15 at 10:00