What I do sometimes is binning the variable into 8 bins, the way that each one of the bins contains an equal portion of the positive targets in the dataset. It's very simple to develop and it give some good and logical results. I'm not sure it will perform that well when modelling rare events.
... View more
