Porovnání metod rekalibrace skóringového modelu kreditního rizika

The main objective of this thesis is to compare effectiveness of the chosen methods of credit risk scoring model (re)calibration and give recommendation on what approach to use. For the comparison, the following methods are selected: intercept shift, intercept and slope shift, their versions extended by the predictions of future portfolio log odds. To fulfill the aim of the work, the thesis is divided into six parts. The first chapter describes prior work in the field of scoring models calibration. The second chapter presents the steps that are taken to construct a probability model and transform it into a scorecard. The third chapter introduces the recalibration approaches that will be employed in the practical part. In the fourth chapter the credit risk model is developed and then converted into the scorecard. In the fifth chapter the recalibration approaches are applied and tested. In the final chapter the conclusions are made. The model of interest is applicable for a portfolio of private clients and is behavioral in nature. The main contribution of this thesis is recommendation on what (re)calibration methods to use in a regular behavioral scoring.