The problem of determining the probability of default or credit risk for companies is crucial when providing financial services. This problem is often modeled based on snapshot data that does not take the time dimension into account. Instead, we approach the problem with enterprise resource planning data in time series. With the added complexity the time series introduce, we pose that deep learning models could be suitable for the task. A comparison of a fully convolutional network and a transformer encoder was made to the current state-of-the-art model for the probability of default problem, XGBoost. The comparison showed that XGBoost generalized very well to the time series domain, even well enough to beat the deep learning models across all evaluation metrics. Furthermore, time series data with monthly, quarterly and yearly timestamps over three years was tested. Also, public features that could be extracted from quarterly and annual financial reports were compared with internal enterprise resource planning data. We found that the introduction of time series to the problem improves the performance and that models based on internal data outperform the ones based on public data. To be more precise, we argue that the dataset being based on small to medium-sized companies lessens the impact of highly granular data, and makes the selection of what features to include more prominent. This is something XGBoost takes advantage of in a very efficient way, especially when extracting features that capture the behavior of the time series, causing it to beat the deep learning competitors even though it does not pick up on the sequential aspect of the data.

Este trabalho tem por objetivo discutir as metodologias de estimação da PD utilizadas na indústria financeira. Além disso, contextualizar a aplicação do trabalho ao IFRS9 e seu direcionamento para o tema de Risco de Crédito. Historicamente os grandes bancos múltiplos utilizam variadas metodologias econométricas para modelar a Probabilidade de Descumprimento (PD),um dos métodos mais tradicionais é a regressão logística, entretanto com a necessidade do cálculo da Perda Esperada de Crédito através do IFRS9, se torna necessário mudar o paradigma de estimação para uma abordagem forward-looking, isto está sendo interpretado por muitas instituições e consultorias como a inclusão de fatores e variáveis projetadas dentro do processo de estimação, ou seja, não serão utilizados apenas os dados históricos para prever o descumprimento ou inadimplência. Dentro deste contexto será proposto uma abordagem que une a estimação da Probabilidade de Descumprimento com a inclusão de um fator foward-looking. / This paper aims to discuss the methodologies used to estimate the Probability Of Default used in the financial industry. In addition, contextualize the application of the work to IFRS9 requirements and its targeting to the Credit Risk theme. Historically large multi-banks use a variety of econometric methodologies to model the Probability of Default, one of the more traditional methods is logistic regression. However, with the need to calculate the expected credit loss through IFRS9, it becomes necessary to change the estimation paradigm to a forwardlooking approach, this is being interpreted by many institutions and consultancies companies as the inclusion of factors and variables projected within the estimation process, that is, not only historical data are used to predict the default. Within this context will be proposed an approach that joins the estimation of Probability of Default with the inclusion of a forward-looking factor.

När propositionen för ett reformerat studiestödssystem lades 1999 poängterades det att studiestödssystemet skulle bära sina egna kostnader. Trots det skrivs stora belopp av. Både Riksrevisionen och Riksgälden har visat att CSN inte använder vedertagna metoder vid beräkningen av det som förväntas gå förlorat på grund av misskötta betalningar. Uppsatsens syfte har varit att skatta vad misskötta betalningar väntas kosta staten i form av framtida avskrivningar samt beräkna vad det skulle innebära för individen att istället bära kostnaden. Som en del i det arbetet har även faktorer som påverkar sannolikheten för misskötta betalningar av studielån identifierats. Resultaten av denna uppsats har bland annat visat att sannolikheten för misskötta betalningar är lägre för individer med eftergymnasial utbildning, hög skuld och låg ålder. Statens kreditförluster på studielån för till exempel individer som blev återbetalningsskyldiga under 2012 förväntas bli mellan 100 och 338 miljoner kronor. Om denna kostnad istället skulle bäras av årskullen innebär det en kostnadsökning på 2,2-7,8 procent för en individ med genomsnittlig skuld.

When banks lend money to another party they face a risk that the borrower will not fulfill its obligation towards the bank. This risk is called credit risk and it’s the largest risk banks faces. According to the Basel accord banks need to have a certain amount of capital requirements to protect themselves towards future financial crisis. This amount is calculated for each loan with an attached risk-weighted asset, RWA. The main parameters in RWA is probability of default and loss given default. Banks are today allowed to use their own internal models to calculate these parameters. Thus hold capital with no gained interest is a great cost, banks seek to find tools to better predict probability of default to lower the capital requirement. Machine learning and supervised algorithms such as Logistic regression, Neural network, Decision tree and Random Forest can be used to decide credit risk. By training algorithms on historical data with known results the parameter probability of default (PD) can be determined with a higher certainty degree compared to traditional models, leading to a lower capital requirement. On the given data set in this article Logistic regression seems to be the algorithm with highest accuracy of classifying customer into right category. However, it classifies a lot of people as false positive meaning the model thinks a customer will honour its obligation but in fact the customer defaults. Doing this comes with a great cost for the banks. Through implementing a cost function to minimize this error, we found that the Neural network has the lowest false positive rate and will therefore be the model that is best suited for this specific classification task. / När banker lånar ut pengar till en annan part uppstår en risk i att låntagaren inte uppfyller sitt antagande mot banken. Denna risk kallas för kredit risk och är den största risken en bank står inför. Enligt Basel föreskrifterna måste en bank avsätta en viss summa kapital för varje lån de ger ut för att på så sätt skydda sig emot framtida finansiella kriser. Denna summa beräknas fram utifrån varje enskilt lån med tillhörande risk-vikt, RWA. De huvudsakliga parametrarna i RWA är sannolikheten att en kund ej kan betala tillbaka lånet samt summan som banken då förlorar. Idag kan banker använda sig av interna modeller för att estimera dessa parametrar. Då bundet kapital medför stora kostnader för banker, försöker de sträva efter att hitta bättre verktyg för att uppskatta sannolikheten att en kund fallerar för att på så sätt minska deras kapitalkrav. Därför har nu banker börjat titta på möjligheten att använda sig av maskininlärningsalgoritmer för att estimera dessa parametrar. Maskininlärningsalgoritmer såsom Logistisk regression, Neurala nätverk, Beslutsträd och Random forest, kan användas för att bestämma kreditrisk. Genom att träna algoritmer på historisk data med kända resultat kan parametern, chansen att en kund ej betalar tillbaka lånet (PD), bestämmas med en högre säkerhet än traditionella metoder. På den givna datan som denna uppsats bygger på visar det sig att Logistisk regression är den algoritm med högst träffsäkerhet att klassificera en kund till rätt kategori. Däremot klassifiserar denna algoritm många kunder som falsk positiv vilket betyder att den predikterar att många kunder kommer betala tillbaka sina lån men i själva verket inte betalar tillbaka lånet. Att göra detta medför en stor kostnad för bankerna. Genom att istället utvärdera modellerna med hjälp av att införa en kostnadsfunktion för att minska detta fel finner vi att Neurala nätverk har den lägsta falsk positiv ration och kommer därmed vara den model som är bäst lämpad att utföra just denna specifika klassifierings uppgift.

Submitted by Diego Peterlevitz Frota (theairguitar@yahoo.com.br) on 2015-09-18T04:44:47Z No. of bitstreams: 1 Dissertação de Mestrado v2.2.1 FINAL.pdf: 702750 bytes, checksum: c584d69fa5b1e9246d10622f4fad5e64 (MD5) / Approved for entry into archive by Renata de Souza Nascimento (renata.souza@fgv.br) on 2015-09-22T13:51:12Z (GMT) No. of bitstreams: 1 Dissertação de Mestrado v2.2.1 FINAL.pdf: 702750 bytes, checksum: c584d69fa5b1e9246d10622f4fad5e64 (MD5) / Made available in DSpace on 2015-09-22T14:21:01Z (GMT). No. of bitstreams: 1 Dissertação de Mestrado v2.2.1 FINAL.pdf: 702750 bytes, checksum: c584d69fa5b1e9246d10622f4fad5e64 (MD5) Previous issue date: 2015-08-24 / This paper covers the fundamentals of the relation between the risk-neutral measure and the real-world, exhibiting some known methods of transforming probability measure associated with each of these two contexts. We show how bonds can be used to estimate the probability of default by their issuers, explaining the reasons that cause it does not reflect, at first, the data observed historically. Using data from Brazilian companies, we estimate the ratio between the risk-neutral and actual probability of default. These results, when compared with other similar studies suggest that the risk premium of Brazilian companies is higher than that of American companies. / Este trabalho aborda os fundamentos da relação entre a medida neutra a risco e o mundo físico, apresentando algumas metodologias conhecidas de transformação da medida de probabilidade associada a cada um destes dois contextos. Mostramos como titulos de crédito podem ser utilizados para a estimação da probabilidade de inadimplência de seus emissores, explicitando os motivos que fazem com que ela não reflita, em um primeiro momento, os dados observados historicamente. Utilizando dados de empresas brasileiras, estimamos a razão entre a probabilidade de default neutra a risco e a probabilidade de default real. Tais resultados, quando comparados com outros trabalhos similares, sugerem que a razão do prêmio de risco de empresas brasileiras possui valor maior do que a de empresas americanas.

Este trabalho tem por objetivo discutir as metodologias de estimação da PD utilizadas na indústria financeira. Além disso, contextualizar a aplicação do trabalho ao IFRS9 e seu direcionamento para o tema de Risco de Crédito. Historicamente os grandes bancos múltiplos utilizam variadas metodologias econométricas para modelar a Probabilidade de Descumprimento (PD),um dos métodos mais tradicionais é a regressão logística, entretanto com a necessidade do cálculo da Perda Esperada de Crédito através do IFRS9, se torna necessário mudar o paradigma de estimação para uma abordagem forward-looking, isto está sendo interpretado por muitas instituições e consultorias como a inclusão de fatores e variáveis projetadas dentro do processo de estimação, ou seja, não serão utilizados apenas os dados históricos para prever o descumprimento ou inadimplência. Dentro deste contexto será proposto uma abordagem que une a estimação da Probabilidade de Descumprimento com a inclusão de um fator foward-looking. / This paper aims to discuss the methodologies used to estimate the Probability Of Default used in the financial industry. In addition, contextualize the application of the work to IFRS9 requirements and its targeting to the Credit Risk theme. Historically large multi-banks use a variety of econometric methodologies to model the Probability of Default, one of the more traditional methods is logistic regression. However, with the need to calculate the expected credit loss through IFRS9, it becomes necessary to change the estimation paradigm to a forwardlooking approach, this is being interpreted by many institutions and consultancies companies as the inclusion of factors and variables projected within the estimation process, that is, not only historical data are used to predict the default. Within this context will be proposed an approach that joins the estimation of Probability of Default with the inclusion of a forward-looking factor.

The diploma thesis focuses on issuance of the primary bond by non-financial companies on the Prague Stock Exchange (PSE). In the theoretical part were described the main parameters of securities and financial indicators of companies that build the risk premium and discussed options of risk-free base. The application part presents the evaluation of major factors influencing price and bond rates on the example of emissions carried on the PSE.

During the recent years, numerous so-called Buy Now, Pay Later companies have emerged. A type of financial institution offering short term consumer credit contracts. As these institutions have gained popularity, their undertaken credit risk has increased vastly. Simultaneously, the IFRS 9 regulatory requirements must be complied with. Specifically, the Probability of Default (PD) for the entire lifetime of such a contract must be estimated. The collection of incremental PDs over the entire course of the contract is called the PD term structure. Accurate estimates of the PD term structures are desirable since they aid in steering business decisions based on a given risk appetite, while staying compliant with current regulations. In this thesis, the efficiency of Machine Learning within PD term structure modeling is examined. Two categories of Machine Learning algorithms, in five variations each, are evaluated; (1) Deep Neural Networks; and (2) Gradient Boosted Trees. The Machine Learning models are benchmarked against a traditional Markov Chain model. The performance of the models is measured by a set of calibration and discrimination metrics, evaluated at each time point of the contract as well as aggregated over the entire time horizon. The results show that Machine Learning can be used efficiently within PD term structure modeling. The Deep Neural Networks outperform the Markov Chain model in all performance metrics, whereas the Gradient Boosted Trees are better in all except one metric. For short-term predictions, the Machine Learning models barely outperform the Markov Chain model. For long-term predictions, however, the Machine Learning models are superior. / Flertalet s.k. Köp nu, betala senare-företag har växt fram under de senaste åren. En sorts finansiell institution som erbjuder kortsiktiga konsumentkreditskontrakt. I samband med att dessa företag har blivit alltmer populära, har deras åtagna kreditrisk ökat drastiskt. Samtidigt måste de regulatoriska kraven ställda av IFRS 9 efterlevas. Specifikt måste fallisemangsrisken för hela livslängden av ett sådant kontrakt estimeras. Samlingen av inkrementell fallisemangsrisk under hela kontraktets förlopp kallas fallisemangsriskens terminsstruktur. Precisa estimat av fallisemangsriskens terminsstruktur är önskvärda eftersom de understödjer verksamhetsbeslut baserat på en given riskaptit, samtidigt som de nuvarande regulatoriska kraven efterlevs. I denna uppsats undersöks effektiviteten av Maskininlärning för modellering av fallisemangsriskens terminsstruktur. Två kategorier av Maskinlärningsalgoritmer, i fem variationer vardera, utvärderas; (1) Djupa neuronnät; och (2) Gradient boosted trees. Maskininlärningsmodellerna jämförs mot en traditionell Markovkedjemodell. Modellernas prestanda mäts via en uppsättning kalibrerings- och diskrimineringsmått, utvärderade i varje tidssteg av kontraktet samt aggregerade över hela tidshorisonten. Resultaten visar att Maskininlärning är effektivt för modellering av fallisemangsriskens terminsstruktur. De djupa neuronnäten överträffar Markovkedjemodellen i samtliga prestandamått, medan Gradient boosted trees är bättre i alla utom ett mått. För kortsiktiga prediktioner är Maskininlärningsmodellerna knappt bättre än Markovkedjemodellen. För långsiktiga prediktioner, däremot, är Maskininlärningsmodellerna överlägsna.

In recent years, the Buy Now Pay Later (BNPL) consumer credit industry associated with e-commerce has been rapidly emerging as an alternative to credit cards and traditional consumer credit products. In parallel, the regulation IFRS 9 was introduced in 2018 requiring creditors to become more proactive in forecasting their Expected Credit Losses and include the impact of macroeconomic factors. This study evaluates several methods of supervised statistical learning to model the Probability of Default (PD) for BNPL credit contracts. Furthermore, the study analyzes to what extent macroeconomic factors impact the prediction under the requirements in IFRS 9 and was carried out as a case study with the Swedish fintech firm Klarna. The results suggest that XGBoost produces the highest predictive power measured in Precision-Recall and ROC Area Under Curve, with ROC values between 0.80 and 0.91 in three modeled scenarios. Moreover, the inclusion of macroeconomic variables generally improves the Precision-Recall Area Under Curve. Real GDP growth, housing prices, and unemployment rate are frequently among the most important macroeconomic factors. The findings are in line with previous research on similar industries and contribute to the literature on PD modeling in the BNPL industry, where limited previous research was identified. / De senaste åren har Buy Now Pay Later (BNPL) snabbt vuxit fram som ett alternativ till kreditkort och traditionella kreditprodukter, i synnerhet inom e-handel. Dessutom introducerades 2018 det nya regelverket IFRS 9, vilket kräver att banker och andra kreditgivare ska bli mer framåtblickande i modelleringen av sina förväntade kreditförluster, samt ta hänsyn till effekter från makroekonomiska faktorer. I denna studie utvärderas flera metoder inom statistisk inlärning för att modellera Probability of Default (PD), sannolikheten att en kreditförlust inträffar, för BNPL-kreditkontrakt. Dessutom analyseras i vilken utsträckning makroekonomiska faktorer påverkar modellernas prediktiva förmågor enligt kraven i IFRS 9. Studien genomfördes som en fallstudie med det svenska fintechföretaget Klarna. Resultaten tyder på att XGBoost har den största prediktionsförmågan mätt i Precision-Recall och ROC Area Under Curve, med ROC-värden mellan 0.80 och 0.91 i tre scenarier. Inkludering av makroekonomiska variabler förbättrar generellt PR-Area Under Curve. Real BNP-tillväxt, bostadspriser och arbetslöshet återfinns frekvent bland de viktigaste makroekonomiska faktorerna. Resultaten är i linje med tidigare forskning inom liknande branscher och bidrar till litteraturen om att modellera PD i BNPL-branschen där begränsad tidigare forskning hittades.

The number of statistical and mathematical credit risk models that financial institutions use and manage due to international and domestic regulatory pressures in recent years has steadily increased. This thesis examines the evolution of model risk management and provides some guidance on how to effectively build and manage different bagging and boosting machine learning techniques for estimating expected credit losses. It examines the pros and cons of these machine learning models and benchmarks them against more conventional models used in practice. It also examines methods for improving their interpretability in order to gain comfort and acceptance from auditors and regulators. To the best of this author’s knowledge, there are no academic publications which review, compare, and provide effective model risk management guidance on these machine learning techniques with the purpose of estimating expected credit losses. This thesis is intended for academics, practitioners, auditors, and regulators working in the model risk management and expected credit loss forecasting space. / Dissertation / Doctor of Philosophy (PhD)