magistrsko delo
Abstract
Modeliranje tveganja je za finančne institucije izjemnega pomena, saj imajo močno spodbudo, da se morajo zaščititi pred neizpolnitvijo obveznosti in obenem zadostiti številnim regulatornim zahtevam. Za boljše razumevanje tematike smo v uvodu povzeli tržno, operativno in kreditno tveganje, s poudarkom na slednjem. To delo se osredotoča na raziskavo verjetnosti neplačila v maloprodajnem bančnem sektorju z uporabo več matematičnih pristopov: statističnih modelov, verjetnostnih modelov, Bayesove statistike, strojnega učenja in klasifikacijske analize. Natančneje smo izbrali modele z različnim matematičnim ozadjem, ki so se najbolj prilagajali realnim podatkom: Mertonov model, linearna in kvadratna diskriminacijska analiza, metoda pomožnih vektorskih klasifikatorjev (SVM) in metoda markovskih verig. Zaključili smo z empirično analizo, kjer smo pod privzetkom začetnih predpostavk preizkusili vsak model in izvedli temeljito analizo, razpravljali o prednostih in slabostih ter predlagali potencialne izboljšave.
Keywords
analiza kreditnega tveganja;diskriminantna analiza;verjetnost neplačila;Mertonov model;metoda pomožnih vektorjev;markovske verige;klasifikacija;
Data
Language: |
Slovenian |
Year of publishing: |
2018 |
Typology: |
2.09 - Master's Thesis |
Organization: |
UL FMF - Faculty of Mathematics and Physics |
Publisher: |
[M. Krković] |
UDC: |
519.2 |
COBISS: |
18459993
|
Views: |
786 |
Downloads: |
275 |
Average score: |
0 (0 votes) |
Metadata: |
|
Other data
Secondary language: |
English |
Secondary title: |
Modelling Probability of Default in the banking sector |
Secondary abstract: |
Modelling of risk is of paramount importance for financial institutions, as they have strong incentive to be protected against default events and are required to comply with numerous regulations. For better understanding we introduce a global overview across market, operational and credit risk with putting a focus on former one. This paper makes an attempt to investigate probability of default in retail banking sector using several mathematical approaches: statistical models, probabilistic models, bayesian statistics, machine learning and classification analysis. More specifically, we selected models from various different domains which are tailored to best model real world data: Merton model, Linear and Quadratic Discriminant Analysis, Support Vector Machines (SVMs) and Markov Chains. We will conclude with an empirical analysis to test each model under certain initial assumptions and conduct a thorough evaluation, discussing advantages and disadvantages of each model and suggesting potential improvements. |
Secondary keywords: |
credit risk analysis;discriminant analysis;probability of default;Merton model;support vector machine;Markov chains;classification; |
Type (COBISS): |
Master's thesis/paper |
Study programme: |
0 |
Embargo end date (OpenAIRE): |
1970-01-01 |
Thesis comment: |
Univ. v Ljubljani, Fak. za matematiko in fiziko, Oddelek za matematiko, Finančna matematika - 2. stopnja |
Pages: |
50 str. |
ID: |
10959809 |