Kádár, Ferenc (OTP Bank)
Advanced
statistical methods for credit risk modeling in practice
In the past 10-15 years logistic
regression became best pracice for probability of default (PD) modeling in commercial
banks. The aim of this presentation is to present pros and cons between this
relatively simple technique and alternative modeling methods such as gradient
boosting, random forests and neural networks. We will also discuss practical
issues such as handling missing values, low default rates, low number of
observations.
Date: Sep. 27, Tuesday 4:15pm
Place: BME, Building „Q”, Room QBF13