The final and most important transformation is quantile binning. The goal with quantile binning is to transform a numeric variable into a categorical one in order to better extract the relation between the variable and the prediction target. This is particularly useful in the presence of nonlinearities between a variable and the target. By splitting the original numeric variables values into n bins of equal size, it is possible to substitute each value by a corresponding bin. Since the number of bins is finite (from 2 to 1,000), the variable is now categorical. Syntax is quantile_bin(var, N) with N the number of bins.
There are two types of unsupervised binning, equal frequency and equal width binning. In equal frequency, ...
