In Predictive Modeling Comparisons: Can you guess why a company might select the best model based on cumulative gains at 30%?
A predictive model is said to be a best fit, if
1. Top 30% of the cumulative gains capture the maximum proportion of the target group
2. Separation of Target vs Non-target should be best at top 30% i.e, Kolmogrov Smirnov Statistic is maximum in 3rd decile (upto which 30% of the total population is captured)
So, a company might select the best model which captures maximum cumulative gains at 30%
In Predictive Modeling Comparisons: Can you guess why a company might select the best model based...