Course Overview
The course will provide attendees a comprehensive overview of Machine Learning techniques applied to credit risk modelling. A hands-on approach is followed by providing both the theoretical and practical toolkit to use on a day-by-day basis. The open-source statistical software R paves the way for grasping all details required to create customized analysis.
During the first day, the key instruments used for modelling are explored. A wide use of the software R characterizes the course from the very beginning. In day one, the emphasis is on familiarizing with Machine Learning techniques and R programming. Indeed, an extensive interaction with R paves the way for the next three-day program.
The focus of the second day is on PD modelling. Starting from an introduction to scorecard one-year modelling, a series of Machine Learning techniques is introduces. Focus is mainly on classification and regression trees, bagging, boosting, and random forest. Hints are also provided on reinforcement learning. A time horizon expansion to encompass the entire lifetime characterizes the second part of the day where survival analysis is introduced and a combination of machine learning techniques is explored by means of R software.
The third day covers both EAD and LGD modelling. A series of approaches is investigated by means of machine learning techniques studied during Day 1 and Day 2. Behavioural model encompassing prepayment, overpayment and a comprehensive EAD dynamic are studied through the lenses of bagging, boosting and random forest modelling. Similarly, LGD is explored by considering both traditional approaches like logit, tobit models as well as through Machine Learning toolkit.
Learning Objectives
- Working-level knowledge of modelling and corresponding hands-on R software development.
- Advanced knowledge of classification and regression trees, bagging, boosting, random forest, and introductory knowledge of reinforcement learning.
- Working knowledge of one-year and lifetime PD modelling based on machine learning techniques.
- Working knowledge of EAD and LGD modelling via machine learning.
