Athens University Economics and Business (AUEB)
Applications of Markov chains and convex sets in ridk management
Abstract
dc:descriptionIn this dissertation we examine the conditions under which a model with a large number of elements can be approximated by a new, simpler one, consisting of a smaller number of elements, without missing the qualitative features of the initial model. The certain conditions facilitate the analysis of large and complex systems that are widely used in studies. We apply our research in two important tools for credit risk management: credit ratings and credit scoringParticularly, we first introduce Markov Chains as a model of the stochastic evolution of credit ratings and then consider the conditions under which a Markov chain is lumpable. We review briefly the definition and characterization of an exactly lumpable Markov chain and through an example of a credit migration matrix we show that in general these conditions are not satisfied. We then introduce the concept of approximate lumpability, and we propose a procedure for finding a lumpable Markov chain that is the closest approximation of the original (non lumpable) Markov chain, in an appropriate norm.We also consider the question of when the existence of a credit scoring system is possible, i.e. provide a very simple test as to whether for a given set of customers, a reliable scoring rule can exist, and if so, provide a candidate for that. Our approach, which is based on a theorem of separating hyperplanes, due to Kirchberger, allows to answer this question by working with a minimal subset of the borrowers data base, thus leading to a very fast and reliable test on whether a scoring system exists or not. Our method is illustrated by a fully worked example
Degree
thesis:*- Grantor dc:publisher
- Athens University Economics and Business (AUEB)
- Year dc:date
- 2014
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Loizides, Marios
- Λοϊζίδης, Μάριος
Subjects
dc:subject × 20- Αλυσίδες Markov
- Γραμμικός διαχωρισμός υπερσυνόλων
- Πιστωτική κατάταξη
- Πιστωτική βαθμολόγηση
- Συσσώρευση αλυσίδων Markov
- Κυρτά σύνολα
- Θεώρημα Kirchberger
- Προσεγγιστική συσσώρευση αλυσίδων Markov
- Markov chains
- Linear separation of hyperplanes
- Credit rating
- Credit scoring
- Lumping Markov chains
- Convex sets
- Kirchberger's theorem
- Approximate lumpability
- Κοινωνικές Επιστήμες
- Οικονομικά και Επιχειρήσεις
- Social Sciences
- Economics and Business
Rights
- Language dc:language
- eng
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/34686