Dissertations / Theses on the topic 'Behaviour scoring'

This thesis stems from my research into the broad area of (credit) scoring and the predicting, understanding and explaining of consumer behaviour. This research started at the Univers1ty of Edinburgh on an ESRC funded project in 1988. This work, which is being submitted as the partial fulfilment of the requirements for the award of Doctor of Philosophy of Loughborough Unvers1ty, consists of an introductory chapter and a selection of papers published 1991 - 2001 (inclusive). The papers address some of the key issues and areas of interest and concern arising from the rapidly evolving and expanding credit (card) market and the highly competitive nature of the credit industry. These features were particularly evident during the late 1980's and throughout the 90's Chapter One provides a general background to the research and outlines some of the key (practical) issues involved in building a (credit) scorecard Additionally, it provides a brief summary of each of the research papers appearing in full in Chapters 2- 9 (inclusive) and ends with some general limitations and conclusions. The research papers appearing in Chapters 2-9 inclusive) are all concerned with predicting, understanding and explaining different types of consumer behaviour in relation to the use of credit cards. For example discriminating between 'GOOD' and 'BAD' repayers of credit card debt on the basis of different definitions of good and bad, the identification of 'slow payers' using different statistical methods; examining the characteristics of credit card users and non-users, and identifying the characteristics of credit card holders most likely to return their credit card.

Made available in DSpace on 2016-04-25T18:39:35Z (GMT). No. of bitstreams: 1 Wagner Schmidt.pdf: 8727051 bytes, checksum: 9669cd75306633dfdd1a2d712ce4d2a3 (MD5) Previous issue date: 2010-10-28<br>This study is the result of the present observation of the movement of civil lawsuits that are growing every day on the market of financial institutions. Nowadays, especially in financial institutions, significant civil lawsuits has been a concern of executives. The main objective of this study is to propose a model of risk management for contingent liabilities in financial institutions, since the difficulty of managing such numbers in the deal result. This is an adaptation of the instruments used in the management of credit risk for the legal area. The models used are the Behaviour Scoring and Credit Scoring. The first model is based on the curve behavioral processes, in this work are denominated like variables. These variables are known industry products offered by financial institutions. On a second level is taken into account the reasons, known as triggering events that led to the civil suits. The second model, Credit Scoring, based on a statistical study of values, which serve as the basis in determining the historical losses. The proposed study is to assist the risk management of these liabilities, eliminating the subjectivity of analysis and allowing greater speed in information. The present results prove that it is possible to use the instruments in question to the risk management of contingent liabilities, reducing the subjectivity of analysis, as greater adherence to criteria and faster responses for managers. The top ten products analyzed shows the results of Credit Scores, for the respective taxable events, termed here as Behaviour Scores. This work, in addition to demonstrating the applicability of the models Credit Scoring and Behavior Scoring also allows us to expand this study to other fields of activities, such as telecommunications, energy, companies that handle large volumes of civil lawsuits, as well as expanded discussion of risk allocation of contingent liability for the product<br>Este estudo é o resultado da observação atual do movimento de ações cíveis que vem crescendo a cada dia no mercado de instituições financeiras. Nos dias atuais, principalmente nas instituições financeiras, volumes significativos de ações judiciais cíveis tem sido motivo de preocupação dos executivos. O principal objetivo deste estudo é propor um modelo de gestão de risco para passivos contingentes nas instituições financeiras, visto a dificuldade de gestão desses números dentro do resultado do negócio. Trata-se de uma adaptação dos instrumentos utilizados na área de gestão de risco de crédito para a área jurídica. Os modelos utilizados em questão são o Behaviour Scoring e o Credit Scoring. O primeiro modelo baseia-se na curva comportamental dos processos, que neste trabalho denominam-se como variáveis. Estas variáveis são os conhecidos produtos ofertados pela indústria das instituições financeiras. Em um segundo nível é levado em consideração os motivos, ou seja, fatos geradores que geraram as ações cíveis. O segundo modelo, o Credit Scoring, baseia-se em um estudo estatístico de valores, os quais servirão de base na apuração das perdas históricas. A proposta do estudo é auxiliar a gestão do risco desses passivos, eliminando a subjetividade de análise e permitindo maior velocidade nas informações. Os resultados obtidos neste trabalho provam que é possível utilizar os instrumentos em questão para a gestão do risco do passivo contingente, diminuindo a subjetividade de análise, visto maior aderência nos critérios e respostas mais rápidas para os gestores. O top ten de produtos analisados mostra os resultados dos Credit Scores, para os respectivos fatos geradores, denominado neste trabalho como Behaviour Scores. Este trabalho, além de evidenciar a aplicabilidade dos modelos Credit Scoring e Behaviour Scoring, também permite expandir este estudo para outros ramos de atividades, como telefonia, energia, empresas que operam com grandes volumes de ações cíveis, além de expandir discussões como alocação de risco de passivo contingente por produto

This thesis will deal with problems and solutions of cross-platform developing using MoNo framework as a replacement of Microsoft .NET framework on Linux and Mac OS-X platforms. It will take in account matters such as limitations in the filesystem to problems with deploying released programs. It will also deal with demands of precise timing and the need of efficient code on precise tasks to construct a program used for creating data from recordings of animals. These animals is set to perform a task, for example exploring a labyrinth or running on a rod, and it is all recorded on video. These videos are later reviewed by an observer which transcripts the recordings into data based on predefined behaviors and the time and frequency with which the animal is expressing them.

Submitted by Fabio Sobreira Campos da Costa (fabio.sobreira@ufpe.br) on 2016-07-12T12:11:15Z No. of bitstreams: 2 license_rdf: 1232 bytes, checksum: 66e71c371cc565284e70f40736c94386 (MD5) Tese_Rosalvo_Neto_CIN_2015.pdf: 7674683 bytes, checksum: 99037c704450a9a878bcbe93ab8b392d (MD5)<br>Made available in DSpace on 2016-07-12T12:11:15Z (GMT). No. of bitstreams: 2 license_rdf: 1232 bytes, checksum: 66e71c371cc565284e70f40736c94386 (MD5) Tese_Rosalvo_Neto_CIN_2015.pdf: 7674683 bytes, checksum: 99037c704450a9a878bcbe93ab8b392d (MD5) Previous issue date: 2015-12-11<br>CAPEs<br>A etapa de pré-processamento em um projeto de descoberta do conhecimento é custosa, em geral, consome cerca de 50 a 80% do tempo total de um projeto. É nesta etapa que um banco de dados relacional é transformado para aplicação de um algoritmo de mineração de dados. A transformação dos dados nesta etapa é uma tarefa complexa, uma vez que exige uma forte integração entre projetistas de banco de dados e especialistas do domínio da aplicação. Os frameworks que buscam sistematizar a etapa de transformação dos dados encontrados na literatura apresentam limitações significativas quando aplicados a soluções comportamentais, como Credit Behavior Scoring. Estas soluções visam a auxiliar as instituições financeiras a decidirem sobre a concessão de crédito aos consumidores com base no risco das solicitações. Este trabalho propõe um framework baseado no Desenvolvimento Dirigido por Modelos para sistematizar esta etapa em soluções de Credit Behavior Scoring. Ele é composto por um meta-modelo que mapeia os conceitos do domínio e um conjunto de regras de transformações. As três principais contribuições do framework proposto são: 1) aumentar o poder discriminatório da solução, através da construção de novas variáveis que maximizam o conteúdo estatístico da informação do domínio; 2) reduzir o tempo da transformação dos dados através da geração automática de código e 3) permitir que profissionais e pesquisadores de Inteligência Artificial e Estatística realizem a transformação dos dados sem o auxílio de especialistas de Banco de Dados. Para validar o framework proposto, dois estudos comparativos foram realizados. Primeiro, um estudo comparando o desempenho entre os principais frameworks existentes na literatura e o framework proposto foi realizado em duas bases de dados. Uma base de dados de um conhecido benchmark de uma competição internacional organizada pela PKDD, e outra obtida de uma das maiores empresas de varejo do Brasil, que possui seu próprio cartão de crédito. Os frameworks RelAggs e Validação de Múltiplas Visões Baseado em Correção foram escolhidos como representantes das abordagens proposicional e mineração de dados relacional, respectivamente. A comparação foi realizada através do processo de validação cruzada estratificada, para definir os intervalos de confiança para a avaliação de desempenho. Os resultados mostram que o framework proposto proporciona um desempenho equivalente ou superior aos principais framework existentes, medido pela área sob a curva ROC, utilizando uma rede neural MultiLayer Perceptron, K vizinho mais próximos e Random Forest como classificadores, com um nível de confiança de 95%. O segundo estudo verificou a redução de tempo proporcionada pelo framework durante a transformação dos dados. Para isso, sete times compostos por estudantes de uma universidade brasileira mensuraram o tempo desta atividade com e sem o framework proposto. O teste pareado Wilcoxon Signed-Rank mostrou que o framework proposto reduz o tempo de transformação com um nível de confiança de 95%.<br>The pre-processing stage in knowledge discovery projects is costly, generally taking between 50 and 80% of total project time. It is in this stage that data in a relational database are transformed for applying a data mining technique. This stage is a complex task that demands from database designers a strong interaction with experts who have a broad knowledge about the application domain. The frameworks that aim to systemize the data transformation stage have significant limitations when applied to behavior solutions such as the Credit Behavior Scoring solutions. Their goal is help financial institutions to decide whether to grant credit to consumers based on the credit risk of their requests. This work proposes a framework based on the Model Driven Development to systemize this stage in Credit Behavioral Scoring solutions. It is composed by a meta-model which maps the domain concepts and a set of transformation rules. This work has three main contributions: 1) improving the discriminant power of data mining techniques by means of the construction of new input variables, which embed new knowledge for the technique; 2) reducing the time of data transformation using automatic code generation and 3) allowing artificial intelligence and statistics modelers to perform the data transformation without the help of database experts. In order to validate the proposed framework, two comparative studies were conducted. First, a comparative study of performance between the main existing frameworks found in literature and the proposed framework applied to two databases was performed. One database from a known benchmark of an international competition organized by PKDD, and another one obtained from one of the biggest retail companies from Brazil, that has its own private label credit card. The RelAggs and Correlation-based Multiple View Validation frameworks were chosen as representatives of the propositional and relational data mining approaches, respectively. The comparison was carried out through by a 10-fold stratified cross-validation process with ten stratified parts in order to define the confidence intervals. The results show that the proposed framework delivers a performance equivalent or superior to those of existing frameworks, for the evaluation of performance measured by the area under the ROC curve, using a Multilayer Perceptron neural network, k-nearest neighbors and Random Forest as classifiers, with a confidence level of 95%. The second comparative study verified the reduction of time required for data transformation using the proposed framework. For this, seven teams composed by students from a Brazilian university measured the runtime of this stage with and without the proposed framework. The paired Wilcoxon Signed-Rank’s Test showed that the proposed framework reduces the time of data transformation with a confidence level of 95%.

Communication can be divied into two broad areas namely, the verbal and nonverbal levels. While attention has been paid to nonverbal communication in the literature, few studies address the nonverbal communication that takes place in the natural setting of a therapeutic session. The present study provides such a naturalistic study, where the verbal content of actual therapy sessions are integrated with the nonverbal content to yield a holistic view of the session. An ecosystemic epistemology is adopted in this study, and represents a move away from more traditional approaches to nonverbal behaviour which are largely confined to a positivistic framework of thought and design. Symlog Interaction Scoring is employed as a practical method of assisting observers in distinguishing nonverbal behaviours, which are usually perceived unconsciously, and lifting them into consciousness, allowing this infonnation to be integrated with the meanings and hypotheses generated during therapy. By deliberately including descriptions of nonverbal behaviour, the descriptions of therapy were broadened, thereby providing a more holistic approach to therapy.<br>Psychology<br>M.A. (Clinical Psychology)

This study aims to explore the possibility of a financial entity to produce a predicted model of default. The study aims to compares the performance of an existing model, the FICO and an alternative model, based on cluster analysis method with dataset available. A third option is presented for the analyses of default, which it is the junction of both models. This third method can be implemented in two different ways: the two models agreeing with acceptance of the loan or the two models approving the rejection of the loan.

Over the past decade, the explosive growth of the Internet has led to a surge of interest to understand and predict aggregate behavior of large number of people or agents, particularly when they are connected through an underlying network structure. Numerous Internet-based applications have emerged that are as diverse as getting micro-tasks executed through online labor markets (also known as crowd sourcing) to acquiring new skills through massively open online courses (also known as MOOCs). However, there has been a major inadequacy in existing studies with respect to evaluating the impact of strategic behavior of the agents participating in such networks, crowds, and classrooms. The primary focus of this doctoral work is to understand the equilibrium behaviour emerging from these real-world, strategic environments by blending ideas from the areas of game theory, graph theory, and optimization, to derive novel solutions to these new-age economic models. In particular, we investigate the following three research challenges: (1) How do strategic agents form connections with one another? Will it ever happen that strategically stable networks are social welfare maximizing as well? (2) How do we design mechanisms for eliciting truthful feedback about an object (perhaps a new product or service or person) from a crowd of strategic raters? What can we tell about these mechanisms when the raters are connected through a social network? (3) How do we incentivize better participation of instructors and students in online edu-cation forums? Can we recommend optimal strategies to students and instructors to get the best out of these forums?

博士<br>輔仁大學<br>商學研究所博士班<br>104<br>To propose a practical alumni donor classification model, this study applied customer profitability analysis from relational marketing theory as a basis to investigate the classification of alumni groups by donation profitability using data contained in a university database regarding a large number of university alumni. The ultimate purpose of this study was to provide a new direction for university planning and fundraising strategies. The subject of this case study was a private university in Taiwan, and its alumni database was used to construct an alumni donor type analysis model through the application of four data mining techniques, discriminant analysis, artificial neural network, multivariate adaptive regression splines and support vector machines. This study tries to identify important characteristics of classifying four types of alumni groups: lost alumni without profitability, lost alumni with profitability, retained alumni without profitability, and retained alumni with profitability. The study will compare the classification accuracy if the four classification techniques and suitable strategies can be proposed based on the research findings.

碩士<br>輔仁大學<br>應用統計學研究所<br>97<br>Credit scoring and behavioural scoring are statistical application commonly used by banks to decide whether or not to grant a loan to customers. Recent global depressions have led credit-granting institutions to search for more effective ways to attract creditworthy customers and control losses. At the customer management level, these two kinds of risk-scoring methodologies offer an objective approaches to predict delinquency. The recent development of behavioral scorings, differs from traditional scorings, adds more information of the repayment and usage behavior of customers and thus can be used further to forecast long term default probability and to adjust credit limits as parts of for examples the marketing and operational policy. Some classical statistical methods logistic regression was first used and remains the most important methods to build scoring systems. More new methods were proposed based on the idea of statistical learning. A novel statistical learning method called Glmnet (Generalized Linear Models via Elastic-Net) is proposed recently. Glmnet enables the creative idea of lasso to be extended to generalized linear models (GLM). Compared to traditionally GLM, Glmnet can not only provide parameter estimations and variable selections simultaneously but also create a computationally efficient process by an elastic-net algorithm. In our study, Glmnet is first proposed in the modeling of behavioural scorings classification accuracy analysis systems. We also compare the results with ANN (Artificial Neural Network), SVM (Support Vector Machines) and LDA (Linear Discriminant Analysis).

Tese de mestrado em Matemática Aplicada em Economia e Gestão, apresentada à Universidade de Lisboa, através da Faculdade de Ciências, 2014<br>Uma das principais áreas em constante crescimento nas duas últimas décadas é a das Telecomunicações. É também uma área que tem sofrido grandes alterações devido ao aumento das necessidades dos clientes, às alterações das regras do mercado e à feroz concorrência. Assim, e tendo em conta os seus desejos, surge a necessidade de responder com as melhores propostas. A problemática económica presente leva a que a concessão de crédito seja, hoje em dia, uma das principais formas de satisfazer os clientes neste ramo. A tecnologia e os softwares atuais proporcionam a produção de grande volume de dados e variáveis. Algumas das ferramentas possibilitam aos analistas de crédito terem respostas imediatas, com todos os indicadores necessários para avaliar o processo de cada cliente. Contudo, estas técnicas exigem que o utilizador siga determinados parâmetros para que a tarefa de optimização tenha sucesso. De forma a desenvolver modelos estatísticos para cada grupo de clientes, é necessário ter um vasto conhecimento dos dados da Vodafone Portugal para que seja possível detetar o comportamento histórico de cada cliente e, posteriormente, caracterizá-lo através de um score. Esta pontuação não é 100% eficaz uma vez que se tratam de modelos de previsão, sendo que o objetivo pretende identificar corretamente o maior número possível de bons clientes e de maus clientes. Em função dos vários indicadores que surgem no contexto do problema dos modelos de Credit Scoring é possível que, ao longo do trabalho, existam várias pequenas conclusões. O desfecho do estudo pode culminar em possíveis alterações das regras de negócio ao crédito, desde a política de concessão de crédito até à metodologia das ações de cobrança. Os métodos utilizados nesta área são métodos quantitativos de análise de credito conhecidos mundialmente e utilizados nas mais diversas áreas, sendo que as instituições financeiras são as que mais os desenvolvem. Foram utilizadas técnicas estatísticas, com destaque para o modelo de regressão logística. A análise e interpretação dos resultados evidenciaram que o uso deste método é mais eficaz do que o processo usado atualmente, pois permite identificar um maior número de clientes com grande probabilidade de default.<br>One of the fields in constant growing over the last two decade is Telecommunications. It's also an area that has undergone major changes due to the growing customer necessities, the changing market rules and stiff competition. Thus, the need to respond with the best proposals taking into account their desires. The present economical problematic makes credit extension, nowadays, one of the main ways to satisfy customers in this business. Technology and up to date software provide the production of large volumes of data and variables. Some tools allow credit analysts to have immediate answers, with all the necessary indicators to evaluate each customer process. However, these techniques require the user to follow certain parameters for the optimization task is successful. In order to develop statistical models for each group of customers, you must have a vast knowledge of data from Vodafone Portugal, to be able to detect the historical behavior of each customer and later classify him through a score. This score is not 100% effective since they deal with forecasting models, with the purpose of correctly identify as many as possible of good customers and bad customers. Depending on the various indicators that appear in the context of credit scoring models is possible that, throughout the work, there are several small conclusions. The outcome of the study may result in possible changes in business credit rules, since the credit concession policy until the methodology of collection actions. The methods used in this area are quantitative methods of credit analysis known worldwide and used in several areas, being mostly developed by financial institutions. Statistical techniques were used, with emphasis on the logistic regression model. The analyzing and interpreting the results, which showed that the use of this method is more effective than the process currently used, since it identifies a greater number of clients with high probability default.

Determining which students are at-risk of poorer outcomes -- such as dropping out, failing classes, or decreasing standardized examination scores -- has become an important area of both research and practice in K-12 education. The models produced from this type of predictive modeling research are increasingly used by high schools in Early Warning Systems to identify which students are at risk and intervene to support better outcomes. It has become common practice to re-build and validate these detectors, district-by-district, due to different data semantics and various risk factors for students in different districts. As these detectors become more widely used, however, a new challenge emerges in applying these detectors across a broad spectrum of school districts with varying availability of past student data. Some districts have insufficient high-quality past data for building an effective detector. Novel approaches that can address the complex data challenges a new district presents are critical for advancing the field. Using an ensemble-based algorithm, I develop a modeling approach that can generate a useful model for a previously unseen district. During the ensembling process, my approach, District Similarity Ensemble Extrapolation (DSEE), weights districts that are more similar to the Target district more strongly during ensembling than less similar districts. Using this approach, I can predict student-at-risk status effectively for unseen districts, across a range of grade ranges, and achieve prediction goodness but ultimately fails to perform better than the previously published Knowles (2015) and Bowers (2012) EWS models proposed for use across districts.

碩士<br>輔仁大學<br>管理學研究所<br>95<br>The development of application scoring in domestic credit card market has been trending toward maturity and delicacy, but, nevertheless, the credit risk management of current cardholders has been accusing of shortage and neglect. Accordingly, this study aims to construct the behavioral scoring model for practitioners in banking industry using the superior and most accurately techniques among the traditional discriminant analysis, logistic regression, neural networks and two-stage hybrid method for customer classification, each of which employs 46 independent variables associated with the aspects of demography, delinquency, payment, credit utilization and cash utilization to comprehensively profile customers and attempt to capture the patterns of cardholders’ repayment behaviors in history, periodically recognize the their credibility and ultimately predict the current customers’ future reimbursement status ensuring the improvement of credit management measures for financial institutes. Results of this empirical study obtained both by discriminant analysis and logistic regression manifest the significance of 11 selected variables in the proposed behavioral scoring models, such as the average utilization last 3 month, number of times M1+ last 6 month, loan flag, number of loan, maximal credit card time of association, age, percentage of cash advance to credit line, balance transfer card flag, days since last payment, education, average utilization last 6 month. The current paper concludes that the hybrid method outperforms all the rest 3 techniques since its classification accuracy reaches 84.00%, higher than neural network of 83.20%, logistics regression of 80.60% and discriminant analysis of 80.07%, and is suggested to be utilized for the future related analysis.

碩士<br>國立臺北科技大學<br>商業自動化與管理研究所<br>96<br>The competition of .finance industry is more and more vigorous. For most of the banking organizations, credit card business is considered highly important among their consumers businesses. The effectiveness of the credit cards issued directly impact the profits of the banks. Therefore, it becomes one of the most important topics to operate a customer relationship management to consolidate those valuable customers. Industry should be to improve purchasing amount of each customer as the goal and raise interest rates of the Bank. This study will estimate the label of customers from the credit card behavior of usage by using DEA. Then we analyze the weights of DEA by using ICA. It interprets the characteristic of each DMU. Moreover we can use the results to adjust credit lines revolving expenses and .procedures expenses.

Para criar modelos de apoio à gestão de cobranças de clientes numa instituição financeira de crédito, foram estimados modelos de sobrevivência heterogéneos, para prever a duração até dois acontecimentos: (i) registo do primeiro atraso no pagamento das mensalidades do contrato de crédito; e (ii) registo de atrasos superiores a 90 dias – default. Seguiu-se uma abordagem condicional tipo II, utilizando todos os clientes da amostra para estimar a duração até ao primeiro atraso e uma sub-amostra, com os clientes que registaram esse primeiro atraso, para estimar a duração até default. Para cada acontecimento foram testadas as distribuições exponencial, Weibull, log-normal e log-logística, em modelos agregados e de mistura. A duração até ao primeiro incidente (i) foi estimada através de um modelo de sobrevivência com proporção de imunes. Esta proporção resulta de um modelo logístico utilizando o scoring interno como variável concomitante. Para os não imunes considerou-se que a duração t segue uma distribuição log-normal, com variáveis explicativas para os parâmetros µ e σ. A duração entre o primeiro incidente e uma situação de default (ii) estimou-se através de um modelo de sobrevivência de mistura com 3 segmentos, com uma função de ligação logit multinomial e assumindo também que t segue uma distribuição log-normal. Neste segundo modelo apenas foram modelados os pesos do modelo logit, considerando µ e σ constantes. Os modelos de sobrevivência apresentados incluem maioritariamente informação recolhida na altura da originação, aplicáveis igualmente como modelos de profit scoring, estimando o envolvimento na data de default, dado um cash-flow futuro.<br>To create models that support the receivables management in a financial institution, heterogeneous survival models were estimated to predict time until two events: (i) having at least one payment overdue; and (ii) 90 days overdue - default. A conditional 2 approach was followed, using all customers of the sample to estimate time until a first payment overdue. A second model was developed, considering only the sub-sample of clients who experienced the first overdue. The exponential, Weibull, log-normal and log-logistic distributions were tested in estimating the time to each event, in aggregate and mixture models. Time to the first overdue (i) was predicted through a survival analysis with immunes, with a logistic model to estimate probability of immunity, using internal credit scoring as covariate. For the non-immunes, a log-normal function, with covariates for both parameters, μ and σ, was estimated to predict time to first overdue. The time between the first overdue and default (ii) was estimated by survival mixture model with 3 segments, with a multinomial logit link function and assuming that time to default also follows a log-normal distribution. Covariates on the second model were considered on the proportions of the mixture model, setting the parameters μ and σ as constants in each group. The survival models presented in this thesis are estimated with data collected at the beginning of the loan, allowing its application in a profit scoring model, by predicting the exposure at the time the customer enters into a situation of default, given an expected cash-flow.

Dissertation presented as the partial requirement for obtaining a Master's degree in Data Science and Advanced Analytics<br>The decision whom to grant a credit is of utmost importance for financial institutions in order to develop both financially profitable, as well as widely accessible financial products. To do this, companies have to be able to distinguish credit applicants, who are able and likely to pay back their loan, from those, who will be unable or unwilling to do so in the future. To improve this decision in the future, the integration of additional behavioural data into the credit decision is proposed in this thesis. FinTech firms are increasingly moving interactions between financial institutions and their customers from local bank branches into digital environments. This transformation enables companies to gather and analyze a large set of previously unavailable behavioural indicators, which can help estimate an individuals credit default risk. This study presents the transforming market conditions that FinTech firms operate in from a regulatory, technical and behavioural perspective and outlines the key changes that impact the offering of credit products. Additionally, it presents the leading approaches of consumer credit research and leverages their best practices in the creation of a behavioural risk scoring model for a FinTech company. The evaluation of the model shows that the inclusion of behavioural indicators into the credit decision is able to significantly improve the performance of decision tree based credit risk models. Models trained with additional behavioural data are able to outperform the base variable set in all performed tests, when compared using the AUC and Kolmogorov-Smirnov measures, while showing no change when assessed using the Brier-Score.