04538ntm a22005297i 4500 000700852 CZ-PrVSE 20230909171012.0 m d cr n|||||||||| 230909s2023 xr fsbm 000 0 eng d NEZPRACOVANÝ IMPORT ABA006 cze ABA006 ABA006 rda Habarta, Filip ISIS:99500 dis Modelování úvěrového rizika portfolia pomocí vícestavových modelů kombinovaných s metodami analýzy přežití eng Portfolio Credit Risk Modelling Using Multi-state Models Combined with Survival Methods / Filip Habarta 2023 ?? stran : digital, PDF soubor Vedoucí práce: Ivana Malá Disertační práce (Ph.D.)—Vysoká škola ekonomická v Praze. Fakulta informatiky a statistiky, 2023 Obsahuje bibliografii Textový (vysokoškolská kvalifikační práce) Rok obhajoby 2023 Credit risk modelling takes many forms today and often is in close connection with credit pricing. We are focusing on the study of the loan process in detail and follow it over a period of time while knowing its actual state. We want to achieve that by defining a Reduced form model that can predict probabilities for various events of loan delinquency with a special focus on loan default. The proposed approach is to define a multi-state model where the transitional intensities, also called hazards, are estimated using the survival analysis methods which are the basis for the transitional probabilities predictions, i.e., prediction of default. This modelling approach is originating mainly from the biostatistics field and its use in the credit area is at most scarce in the literature. The model is kept simple and is restricted to the intermediate states describing whether the repayments are met on time or not and to absorbing states loan default and full repayment of the loan. Several options for the model variants are discussed with the focus mainly on the models with Markovian property and the semi-parametric Cox model. The final proposed model is time in-homogenous modulated semi-Markov model as our transition intensities vary over time, are based on the time-dependent covariates and relaxes Markov assumption to the form of historical dependence only on the time since the entry to the modelled state. We have built up the model on the transitional hazards from the Cox semi-parametric model, estimated by the Breslow estimator, and discussed various pitfalls of the approach. To estimate the final state occupancy probabilities over time we have used non-parametric Aalen-Johansen estimator. A great advantage of the proposed model is that it can be used as both, a credit scoring model (predicting the probability of default) and a behavioural scoring model (following the loan process through time). We showcase the model construction, estimation, prediction and validation on the publicly available data from peer-to-peer lending platform. Additionally, the total expected loss of the portfolio of loans is calculated, including the estimate of its probability distribution for a single loan, and its estimates for the basic scenarios of the impact of the change in macroeconomic situation. The overall modelling results are satisfying but also provide a lot of space for future follow-up research. Způsob přístupu: Internet statistika [obor disert. práce] disertace fd132024 czenas dissertations eczenas Credit risk Multi-state model Survival analysis P2P lending Malá, Ivana, 1962- mzk2005282656 ths Cipra, Tomáš, 1952- xx0011487 opn Vysoká škola ekonomická v Praze. Fakulta informatiky a statistiky kn20010709399 dgg https://insis.vse.cz/zp/70096/podrobnosti VŠKP v InSIS https://insis.vse.cz/zp/70096 Hlavní práce https://insis.vse.cz/zp/70096/posudek/vedouci Hodnocení vedoucího https://insis.vse.cz/zp/70096/posudek/oponent/79846 Oponentura https://insis.vse.cz/zp/70096/posudek/oponent/79847 Oponentura https://insis.vse.cz/zp/70096/podrobnosti dc:identifier NEPOSILAT VSKP vse70096 230906 70096