Terminology
- KS (Kolmogorov-Smirnov) Statistics - This is a measure to show how far off one distribution is from another. In Gains#, this is calculated as
KS = 100* max abs(F(s | target = 1) - F(s | target = 0))
F(s|condition) is conditional cumulative probability function for a given condition.
The range of KS is between 0 and 100. Larges KS indicates better predictive model.
