Made available in DSpace on 2014-06-12T15:07:00Z (GMT). No. of bitstreams: 2 arquivo1363_1.pdf: 1828701 bytes, checksum: f0a7740303ce331d14c5ebcc9b8eb289 (MD5) license.txt: 1748 bytes, checksum: 8a4605be74aa9ea9d79846c1fba20a33 (MD5) Previous issue date: 2006 / Os modelos de Credit Scoring são modelos quantitativos empregados comumente por instituições financeiras na mensuração e previsão do risco de crédito, possuindo uso consolidado no processo de concessão de crédito destas instituições. Este trabalho de dissertação objetivou avaliar a possibilidade de aplicação de modelos Credit Scoring em uma instituição de microcrédito denominada Fundo Rotativo de Ação da Cidadania Cred Cidadania, situada em Recife (PE). Para isso, foram coletados dados relativos a uma amostra de clientes do Cred Cidadania, e estes dados foram utilizados para desenvolver dois tipos de modelos de Credit Scoring: um de aprovação de crédito e um outro chamado behavioural scoring (escoragem comportamental). As técnicas estatísticas empregadas na construção dos modelos foram análise discriminante e regressão logística. Os modelos obtidos agregaram variáveis como renda líquida do empreendimento, número de parcelas do empréstimo, número de dependentes do cliente, estado civil do cliente, valor do empréstimo, tempo de funcionamento do empreendimento, eficiência do agente de crédito, dentre outras. Algumas variáveis representam atributos que contribuem para o aumento da propensão à inadimplência do solicitante, enquanto outras colaboram para a redução do risco de inadimplência, o que repercute positivamente na sustentabilidade financeira na instituição. Os resultados do estudo demonstraram que os modelos Credit Scoring obtêm desempenho satisfatório quando utilizados na análise de risco de crédito na instituição de microcrédito Cred Cidadania, alcançando um percentual de classificação correta dos clientes de cerca de 80%. Os resultados indicam também que o uso de modelos Credit Scoring fornece subsídios à instituição, auxiliando-a na prevenção e redução da inadimplência e na diminuição dos seus custos operacionais, dois problemas que afetam a sua sustentabilidade financeira

nÃo hà / Sendo a intermediaÃÃo financeira a principal atividade dos bancos, alocando recursos de clientes superavitÃrios a clientes deficitÃrios, à na incerteza quanto ao carÃter e a capacidade de pagamento dos clientes que se estabelece o risco e com ele a necessidade de se buscar novas alternativas para se proteger de perdas potenciais, que podem refletir em menores lucros para as instituiÃÃes. AlÃm da subjetividade dos analistas de crÃdito, o uso de modelos quantitativos, baseados em prÃticas estatÃsticas, economÃtricas e matemÃticas, vÃm cada vez mais se firmando nos mercados como ferramenta de apoio aos gestores de crÃdito na tomada de decisÃo. VÃrios modelos de avaliaÃÃo de risco sÃo adotados pelas instituiÃÃes, modelos de credit scoring, behavioral scoring, sÃo exemplos destes modelos. O modelo de credit scoring tem sido um dos mais usados, em especial para concessÃo de crÃdito a pessoas fÃsicas. Os modelos de credit scoring utilizam tÃcnicas como a anÃlise de discriminantes, programaÃÃo matemÃtica, econometria, redes neurais, entre outras, para atravÃs da anÃlise de caracterÃsticas particulares dos indivÃduos, estabelecer uma mÃtrica de separaÃÃo de bons e maus pagadores, atribuindo probabilidades diferentes de inadimplÃncia aos mesmos. A presente dissertaÃÃo tem como objetivo central analisar os determinantes de inadimplÃncia (pessoa fÃsica), usando uma abordagem economÃtrica com base no modelo Logit. O modelo utilizado foi um modelo para aprovaÃÃo de crÃdito na abertura de conta corrente, partindo de um estudo com uma amostra de 308 observaÃÃes (cadastros pessoas fÃsicas), baseados na experiÃncia real de uma instituiÃÃo financeira, cujo objetivo à atingir uma taxa de aprovaÃÃo de crÃdito tal que a receita mÃdia depois das perdas de emprÃstimos seja maximizada. / In the financial intermediation, banks focus on its main activity, allocating resources from clients with surplus to deficit clients. The uncertainty related to the characteristics or payment capacity of the clients establishes the risk and the need to search for new alternatives to protect the institutions from potential losses, which may reflect on lower profits. Besides the subjective issue of credit analysts, the use of quantitative models, based on statistical, mathematical or econometric practices are becoming an important tool to support credit managers on the decision making process. There are several models of risk evaluation, which are adopted by financial institutions such as the credit scoring and the behavioral scoring models. The credit-scoring model has been widely used, especially on the concession of individual credit. The credit scoring model uses techniques such as discriminant analysis, mathematic programming, econometrics, neural networks, among others, to analyze particular characteristics of individuals where it establishes a metric separation of good and bad payers, therefore providing different nonpayment status to each. This present dissertation has the main objective of analyzing the determinants of nonpayment status (individuals), using an econometric approach based on the Logit model. The model utilized was a model for approval of credit in the opening from the bill shackle, starting from a study with 308 observations (physical registers Persons), based in the real experience of a financial institution, whose objective is he reach a credit approval rate such that the medium prescription after the losses of loans be maximized.

Com a crescente demanda por crédito é muito importante avaliar o risco de cada operação desse tipo. Portanto, ao fornecer crédito a um cliente é necessário avaliar as chances do cliente não pagar o empréstimo e, para esta tarefa, as técnicas de credit scoring são aplicadas. O presente trabalho apresenta uma revisão da literatura de credit scoring com o objetivo de fornecer uma vis~ao geral das várias técnicas empregadas. Além disso, um estudo de simulação computacional é realizado com o intuito de comparar o comportamento de várias técnicas apresentadas no estudo. / Nowadays the increasing amount of bank transactions and the increasing of data storage created a demand for risk evaluation associated with personal loans. It is very important for a company has a very good tools in credit risk evaluation because theses tools can avoid money losses. In this context, it is interesting estimate the default probability for a customers and, the credit scoring techniques are very useful for this task. This work presents a credit scoring literature review with and aim to give a overview covering many techniques employed in credit scoring and, a computational study is accomplished in order to compare some of the techniques seen in this text.

Klarna uses a logistic regression to estimate the probability that an e-store customer will default on its given credit. The logistic regression is a linear statistical model which cannot detect non-linearities in the data. The aim of this project has been to develop a program which can be used to find suitable non-linear interaction-variables. This can be achieved using a Restricted Boltzmann Machine, an unsupervised neural network, whose hidden nodes can be used to model the distribution of the data. By using the hidden nodes as new variables in the logistic regression it is possible to see which nodes that have the greatest impact on the probability of default estimates. The contents of the hidden nodes, corresponding to different parts of the data distribution, can be used to find suitable interaction-variables which will allow the modelling of non-linearities. It was possible to find the data distribution using the Restricted Boltzmann Machine and adding its hidden nodes to the logistic regression improved the model's ability to predict the probability of default. The hidden nodes could be used to create interaction-variables which improve Klarna's internal models used for credit risk estimates. / Klarna använder en logistisk regression för att estimera sannolikheten att en e-handelskund inte kommer att betala sina fakturor efter att ha givits kredit. Den logistiska regressionen är en linjär modell och kan därför inte upptäcka icke-linjäriteter i datan. Målet med detta projekt har varit att utveckla ett program som kan användas för att hitta lämpliga icke-linjära interaktionsvariabler. Genom att införa dessa i den logistiska regressionen blir det möjligt att upptäcka icke-linjäriteter i datan och därmed förbättra sannolikhetsestimaten. Det utvecklade programmet använder Restricted Boltzmann Machines, en typ av oövervakat neuralt nätverk, vars dolda noder kan användas för att hitta datans distribution. Genom att använda de dolda noderna i den logistiska regressionen är det möjligt att se vilka delar av distributionen som är viktigast i sannolikhetsestimaten. Innehållet i de dolda noderna, som motsvarar olika delar av datadistributionen, kan användas för att hitta lämpliga interaktionsvariabler. Det var möjligt att hitta datans distribution genom att använda en Restricted Boltzmann Machine och dess dolda noder förbättrade sannolikhetsestimaten från den logistiska regressionen. De dolda noderna kunde användas för att skapa interaktionsvariabler som förbättrar Klarnas interna kreditriskmodeller.

In our master thesis, we compare ten classification algorithms for credit scor- ing. Their prediction performances are measured by six different classification performance measurements. We use a unique P2P lending data set with more than 200,000 records and 23 variables for our classifiers comparison. This data set comes from Lending Club, the biggest P2P lending platform in the United States. Logistic regression, Artificial neural network, and Linear discriminant analysis are the best three classifiers according to our results. Random forest ranks as the fifth best classifier. On the other hand, Classification and regression tree and k-Nearest neighbors are ranked as the worse classifiers in our ranking. 1

Credit scoring is important and rapidly developing discipline. The aim of this thesis is to describe basic methods used for building and interpretation of the credit scoring models with an example of application of these methods for designing such models using statistical software R. This thesis is organized into five chapters. In chapter one, the term of credit scoring is explained with main examples of its application and motivation for studying this topic. In the next chapters, three in financial practice most often used methods for building credit scoring models are introduced. In chapter two, the most developed one, logistic regression is discussed. The main emphasis is put on the logistic regression model, which is characterized from a mathematical point of view and also various ways to assess the quality of the model are presented. The other two methods presented in this thesis are decision trees and Random forests, these methods are covered by chapters three and four. An important part of this thesis is a detailed application of the described models to a specific data set Default using the R program. The final fifth chapter is a practical demonstration of building credit scoring models, their diagnostics and subsequent evaluation of their applicability in practice using R. The appendices include used R code and also functions developed for testing of the final model and code used through the thesis. The key aspect of the work is to provide enough theoretical knowledge and practical skills for a reader to fully understand the mentioned models and to be able to apply them in practice.

No autorizada por el autor para ser publicada a texto completo / El presente trabajo de título tuvo como objetivo desarrollar una herramienta de scoring crediticio dirigida a empresas Latinoamericanas emisoras de títulos con clasificación menor o igual a BB. Actualmente en la plaza local, el desconocimiento que existe en este tipo de instrumentos de renta fija se supedita, en su mayoría, a la compra de fondos elaborados por empresas externas y no al estudio y desarrollo de tecnologías in house, externalizando, de esta forma, el análisis crediticio. Por lo tanto, el interés de desarrollar herramientas que apoyen la toma de decisiones es imperante para instituciones como Asesorías e Inversiones Cruz del Sur que busca, evidentemente, obtener retornos por sobre la competencia. La metodología para el scoring consistió en un estudio de los reportes y recomendaciones de los principales bancos de inversión y compañías de servicios financieros, tanto nacional como internacional, que brindan fondos e investigación de empresas Latinoamericanas y mercados emergentes, de tal manera de crear un universo de las principales métricas que son utilizadas en sus análisis actualmente. De tal universo se derivaron, a través de un estudio de incidencias y juicio de expertos, 5 ratios que otorgaban un diagnóstico de la estructura de deuda y capacidad de cumplir con obligaciones en el corto plazo. Posteriormente, se le asignó a cada métrica un puntaje ajustado al percentil diez de la distribución que presentaba y luego, a través de una descomposición del rendimiento del instrumento, se realizaron ejercicios regresivos (lineal y de panel) que estimaron la importancia de cada métrica en la calibración final. La herramienta fue realizada en lenguaje VBA y su interfaz en Excel, otorgando, además del score crediticio, funcionalidades complementarias que incluyeran información de mercado de los títulos, gráficos y fácil manejo de una base de datos interna con objeto de disminuir tiempos asignados al proceso de manejo de información. El resultado, considerando todas las funcionalidades que abarca, fue una herramienta capaz de otorgar una opinión sobre las circunstancias de un emisor para cubrir sus compromisos financieros sujeta a la limitada posibilidad de automatización de las variables y presentar un punto de partida para el departamento de estudios.

This thesis has investigated two-stage regularized logistic regressions applied on the credit scoring problem. Credit scoring refers to the practice of estimating the probability that a customer will default if given credit. The data was supplied by Klarna AB, and contains a larger number of observations than many other research papers on credit scoring. In this thesis, a two-stage regression refers to two staged regressions were the some kind of information from the first regression is used in the second regression to improve the overall performance. In the best performing models, the first stage was trained on alternative labels, payment status at earlier dates than the conventional. The predictions were then used as input to, or to segment, the second stage. This gave a gini increase of approximately 0.01. Using conventional scorecutoffs or distance to a decision boundary to segment the population did not improve performance. / Denna uppsats har undersökt tvåstegs regulariserade logistiska regressioner för att estimera credit score hos konsumenter. Credit score är ett mått på kreditvärdighet och mäter sannolikheten att en person inte betalar tillbaka sin kredit. Data kommer från Klarna AB och innehåller fler observationer än mycket annan forskning om kreditvärdighet. Med tvåstegsregressioner menas i denna uppsats en regressionsmodell bestående av två steg där information från det första steget används i det andra steget för att förbättra den totala prestandan. De bäst presterande modellerna använder i det första steget en alternativ förklaringsvariabel, betalningsstatus vid en tidigare tidpunkt än den konventionella, för att segmentera eller som variabel i det andra steget. Detta gav en giniökning på approximativt 0,01. Användandet av enklare segmenteringsmetoder så som score-gränser eller avstånd till en beslutsgräns visade sig inte förbättra prestandan.

Despite being governed by several laws aimed at preventing racial inequality in access to housing and credit resources, the mortgage lending market remains a contributor to racial and place-based disparities in homeownership rates, wealth, and access to high-quality community resources. Scholarship has identified persistent disparities in mortgage loan approval rates and subprime lending between white borrowers and those from other racial and ethnic groups, and between white neighborhoods and neighborhoods with high levels of non-white residents. Against this backdrop, the mortgage lending industry is undergoing rapid, technology-driven changes in risk assessment and application processing. Traditional borrower risk-assessment methods including face-to-face discussions between lenders and applicants and the prominent use of FICO credit scores have been replaced or supplemented by automated decision-making tools at a new generation of institutions known as fintech lenders. Little is known about the relationship between lenders using these new tools and the racial and spatial disparities that have long defined the wider mortgage market. Given the well-documented history of discrimination in lending along with findings of technology-driven racial inequality in other economic sectors, fintech lending’s potential for racial discrimination warrants increased scrutiny. This dissertation compares the lending outcomes of traditional and fintech mortgage lenders in the United States depending on applicant and neighborhood racial characteristics. Using data from the Home Mortgage Disclosure Act, an original dataset classifying lenders as fintech or traditional, and an array of complimentary administrative data sources, it provides an assessment of the salience of race and place in the rates at which mortgage loans from each lender type are approved and assigned subprime terms. Results from a series of regression-based quantitative analyses suggest fintech mortgage lenders, like traditional mortgage lenders, approve and deny loans and distribute subprime credit at disparate rates to white borrowers and neighborhoods relative to nonwhite borrowers and neighborhoods. Findings suggest that policymakers and regulators must increase their oversight of fintech lenders, ensuring that further advances in lending technology do not concretize longstanding racial and spatial disparities.

This dissertation is concerned with three problems within the field of optimization for agent--based systems. Firstly, the inverse optimal control problem is investigated for the single-agent system. Given a dynamic process, the goal is to recover the quadratic cost function from the observation of optimal control sequences. Such estimation could then help us develop a better understanding of the physical system and reproduce a similar optimal controller in other applications. Next, problems of optimization over networked systems are considered. A novel differential game approach is proposed for the optimal intrinsic formation control of multi-agent systems. As for the credit scoring problem, an optimal filtering framework is utilized to recursively improve the scoring accuracy based on dynamic network information. In paper A, the problem of finite horizon inverse optimal control problem is investigated, where the linear quadratic (LQ) cost function is required to be estimated from the optimal feedback controller. Although the infinite-horizon inverse LQ problem is well-studied with numerous results, the finite-horizon case is still an open problem. To the best of our knowledge, we propose the first complete result of the necessary and sufficient condition for the existence of corresponding LQ cost functions. Under feasible cases, the analytic expression of the whole solution space is derived and the equivalence of weighting matrices is discussed. For infeasible problems, an infinite dimensional convex problem is formulated to obtain a best-fit approximate solution with minimal control residual, where the optimality condition is solved under a static quadratic programming framework to facilitate the computation. In paper B, the optimal formation control problem of a multi-agent system is studied. The foraging behavior of N agents is modeled as a finite-horizon non-cooperative differential game under local information, and its Nash equilibrium is studied. The collaborative swarming behaviour derived from non-cooperative individual actions also sheds new light on understanding such phenomenon in the nature. The proposed framework has a tutorial meaning since a systematic approach for formation control is proposed, where the desired formation can be obtained by only intrinsically adjusting individual costs and network topology. In contrast to most of the existing methodologies based on regulating formation errors to the pre-defined pattern, the proposed method does not need to involve any information of the desired pattern beforehand. We refer to this type of formation control as intrinsic formation control. Patterns of regular polygons, antipodal formations and Platonic solids can be achieved as Nash equilibria of the game while inter-agent collisions are naturally avoided. Paper C considers the credit scoring problem by incorporating dynamic network information, where the advantages of such incorporation are investigated in two scenarios. Firstly, when the scoring publishment is merely individual--dependent, an optimal Bayesian filter is designed for risk prediction, where network observations are utilized to provide a reference for the bank on future financial decisions. Furthermore, a recursive Bayes estimator is proposed to improve the accuracy of score publishment by incorporating the dynamic network topology as well. It is shown that under the proposed evolution framework, the designed estimator has a higher precision than all the efficient estimators, and the mean square errors are strictly smaller than the Cramér-Rao lower bound for clients within a certain range of scores. / I denna avhandling behandlas tre problem inom optimering för agentbaserade system. Inledningsvis undersöks problemet rörande invers optimal styrning för ett system med en agent. Målet är att, givet en dynamisk process, återskapa den kvadratiska kostnadsfunktionen från observationer av sekvenser av optimal styrning. En sådan uppskattning kan ge ökad förståelse av det underliggande fysikaliska systemet, samt vara behjälplig vid konstruktion av en liknande optimal regulator för andra tillämpningar. Vidare betraktas problem rörande optimering över nätverkssystem. Ett nytt angreppssätt, baserat på differentialspel, föreslås för optimal intrinsisk formationsstyrning av system med fler agenter. För kreditutvärderingsproblemet utnyttjas ett filtreringsramverk för att rekursivt förbättra kreditvärderingens noggrannhet baserat på dynamisk nätverksinformation. I artikel A undersöks problemet med invers optimal styrning med ändlig tidshorisont, där den linjärkvadratiska (LQ) kostnadsfunktionen måste uppskattas från den optimala återkopplingsregulatorn. Trots att det inversa LQ-problemet med oändlig tidshorisont är välstuderat och med flertalet resultat, är fallet med ändlig tidshorisont fortfarande ett öppet problem. Så vitt vi vet presenterar vi det första kompletta resultatet med både tillräckliga och nödvändiga villkor för existens av en motsvarande LQ-kostnadsfunktion. I fallet med lösbara problem härleds ett analytiskt uttryck för hela lösningsrummet och frågan om ekvivalens med viktmatriser behandlas. För de olösbara problemen formuleras ett oändligtdimensionellt konvext optimeringsproblem för att hitta den bästa approximativa lösningen med den minsta styrresidualen. För att underlätta beräkningarna löses optimalitetsvillkoren i ett ramverk för statisk kvadratisk programmering. I artikel B studeras problemet rörande optimal formationsstyrning av ett multiagentsystem. Agenternas svärmbeteende modelleras som ett icke-kooperativt differentialspel med ändlig tidshorisont och enbart lokal information. Vi studerar detta spels Nashjämvikt. Att, ur icke-kooperativa individuella handlingar, härleda ett kollaborativt svärmbeteende kastar nytt ljus på vår förståelse av sådana, i naturen förekommande, fenomen. Det föreslagna ramverket är vägledande i den meningen att det är ett systematiskt tillvägagångssätt för formationsstyrning, där den önskade formeringen kan erhållas genom att endast inbördes justera individuella kostnader samt nätverkstopologin. I motstat till de flesta befintliga metoder, vilka baseras på att reglera felet i formeringen relativt det fördefinierade mönstret, så behöver den föreslagna metoden inte på förhand ta hänsyn till det önskade mönstret. Vi kallar denna typ av formationsstyrning för intrinsisk formationsstyrning. Mönster så som regelbundna polygoner, antipodala formeringar och Platonska kroppar kan uppnås som Nashjämvikter i spelet, samtidigt som kollisioner mellan agenter undviks på ett naturligt sätt. Artikel C behandlar kreditutvärderingsproblemet genom att lägga till dynamisk nätverksinformation. Fördelarna med en sådan integrering undersöks i två scenarier. Då kreditvärdigheten enbart är individberoende utformas ett optimalt Bayesiskt filter för riskvärdering, där observationer från nätverket används för att tillhandahålla en referens för banken på framtida finansiella beslut. Vidare föreslås en rekursiv Bayesisk estimator (stickprovsvariabel) för att förbättra noggrannheten på den skattade kreditvärdigheten genom att integrera även den dynamiska nätverkstopologin. Inom den föreslagna ramverket för tidsutveckling kan vi visa att, för kunder inom ett visst intervall av värderingar, har den utformade estimatorn högre precision än alla effektiva estimatorer och medelkvadrafelet är strikt mindre än den nedre gränsen från Cramér-Raos olikhet. / <p>QC 20190603</p>