renero / class_notebooks Goto Github PK

In the Data Manipulation Classes notebook, you have:

houses.scale()
houses.ensure_normality()

The ensure_normality() function appears not to exist. You probably mean the fix_skewness function.

Note that there is also a skewed_features function, which uses Box-Cox. That function can't be applied, because it only works on positive numbers (hence the Yeo-Johnson used in the fix_skewness function). After applying the StandardScaler, the numerical values will be zero centred, so will range from negative to positive.

Also, though, the PowerTransformer has a standardize option, which seems to apply the StandardScaler anyway. But, that doesn't work as there is a "overflow encountered in multiply" problem, unless the StandardScaler is first applied to the data. Hence, I suppose, why standardize=False.