none

Cette thèse se place dans le cadre des modèles d’apprentissage automatique de classification binaire. Le cas d’application est le scoring de risque de crédit. En particulier, les méthodes proposées ainsi que les approches existantes sont illustrées par des données réelles de Crédit Agricole Consumer Finance, acteur majeur en Europe du crédit à la consommation, à l’origine de cette thèse grâce à un financement CIFRE. Premièrement, on s’intéresse à la problématique dite de “réintégration des refusés”. L’objectif est de tirer parti des informations collectées sur les clients refusés, donc par définition sans étiquette connue, quant à leur remboursement de crédit. L’enjeu a été de reformuler cette problématique industrielle classique dans un cadre rigoureux, celui de la modélisation pour données manquantes. Cette approche a permis de donner tout d’abord un nouvel éclairage aux méthodes standards de réintégration, et ensuite de conclure qu’aucune d’entre elles n’était réellement à recommander tant que leur modélisation, lacunaire en l’état, interdisait l’emploi de méthodes de choix de modèles statistiques. Une autre problématique industrielle classique correspond à la discrétisation des variables continues et le regroupement des modalités de variables catégorielles avant toute étape de modélisation. La motivation sous-jacente correspond à des raisons à la fois pratiques (interprétabilité) et théoriques (performance de prédiction). Pour effectuer ces quantifications, des heuristiques, souvent manuelles et chronophages, sont cependant utilisées. Nous avons alors reformulé cette pratique courante de perte d’information comme un problème de modélisation à variables latentes, revenant ainsi à une sélection de modèle. Par ailleurs, la combinatoire associée à cet espace de modèles nous a conduit à proposer des stratégies d’exploration, soit basées sur un réseau de neurone avec un gradient stochastique, soit basées sur un algorithme de type EM stochastique.Comme extension du problème précédent, il est également courant d’introduire des interactions entre variables afin, comme toujours, d’améliorer la performance prédictive des modèles. La pratique classiquement répandue est de nouveau manuelle et chronophage, avec des risques accrus étant donnée la surcouche combinatoire que cela engendre. Nous avons alors proposé un algorithme de Metropolis-Hastings permettant de rechercher les meilleures interactions de façon quasi-automatique tout en garantissant de bonnes performances grâce à ses propriétés de convergence standards. La dernière problématique abordée vise de nouveau à formaliser une pratique répandue, consistant à définir le système d’acceptation non pas comme un unique score mais plutôt comme un arbre de scores. Chaque branche de l’arbre est alors relatif à un segment de population particulier. Pour lever la sous-optimalité des méthodes classiques utilisées dans les entreprises, nous proposons une approche globale optimisant le système d’acceptation dans son ensemble. Les résultats empiriques qui en découlent sont particulièrement prometteurs, illustrant ainsi la flexibilité d’un mélange de modélisation paramétrique et non paramétrique. Enfin, nous anticipons sur les futurs verrous qui vont apparaître en Credit Scoring et qui sont pour beaucoup liés la grande dimension (en termes de prédicteurs). En effet, l’industrie financière investit actuellement dans le stockage de données massives et non structurées, dont la prochaine utilisation dans les règles de prédiction devra s’appuyer sur un minimum de garanties théoriques pour espérer atteindre les espoirs de performance prédictive qui ont présidé à cette collecte. / This manuscript deals with model-based statistical learning in the binary classification setting. As an application, credit scoring is widely examined with a special attention on its specificities. Proposed and existing approaches are illustrated on real data from Crédit Agricole Consumer Finance, a financial institute specialized in consumer loans which financed this PhD through a CIFRE funding. First, we consider the so-called reject inference problem, which aims at taking advantage of the information collected on rejected credit applicants for which no repayment performance can be observed (i.e. unlabelled observations). This industrial problem led to a research one by reinterpreting unlabelled observations as an information loss that can be compensated by modelling missing data. This interpretation sheds light on existing reject inference methods and allows to conclude that none of them should be recommended since they lack proper modelling assumptions that make them suitable for classical statistical model selection tools. Next, yet another industrial problem, corresponding to the discretization of continuous features or grouping of levels of categorical features before any modelling step, was tackled. This is motivated by practical (interpretability) and theoretical reasons (predictive power). To perform these quantizations, ad hoc heuristics are often used, which are empirical and time-consuming for practitioners. They are seen here as a latent variable problem, setting us back to a model selection problem. The high combinatorics of this model space necessitated a new cost-effective and automatic exploration strategy which involves either a particular neural network architecture or Stochastic-EM algorithm and gives precise statistical guarantees. Third, as an extension to the preceding problem, interactions of covariates may be introduced in the problem in order to improve the predictive performance. This task, up to now again manually processed by practitioners and highly combinatorial, presents an accrued risk of misselecting a “good” model. It is performed here with a Metropolis Hastings sampling procedure which finds the best interactions in an automatic fashion while ensuring its standard convergence properties, thus good predictive performance is guaranteed. Finally, contrary to the preceding problems which tackled a particular scorecard, we look at the scoring system as a whole. It generally consists of a tree-like structure composed of many scorecards (each relative to a particular population segment), which is often not optimized but rather imposed by the company’s culture and / or history. Again, ad hoc industrial procedures are used, which lead to suboptimal performance. We propose some lines of approach to optimize this logistic regression tree which result in good empirical performance and new research directions illustrating the predictive strength and interpretability of a mix of parametric and non-parametric models. This manuscript is concluded by a discussion on potential scientific obstacles, among which the high dimensionality (in the number of features). The financial industry is indeed investing massively in unstructured data storage, which remains to this day largely unused for Credit Scoring applications. Doing so will need statistical guarantees to achieve the additional predictive performance that was hoped for.

published_or_final_version / Psychology / Doctoral / Doctor of Philosophy

Modelos de customer scoring são utilizados para mensurar o risco de crédito de clientes de instituições financeiras. Neste trabalho, são apresentadas três estratégias que podem ser utilizadas para o desenvolvimento desses modelos. São discutidas as vantagens de cada uma dessas estratégias, bem como os modelos e a teoria estatística associada a elas. Algumas medidas de performance usualmente utilizadas na comparação de modelos de risco de crédito são descritas. Modelos para cada uma das estratégias são ajustados utilizando-se dados reais obtidos de uma instituição financeira. A performance das estratégias para esse conjunto de dados é comparada a partir de medidas usualmente utilizadas na avaliação de modelos de risco de crédito. Uma simulação também é desenvolvida com o propósito de comparar o desempenho das estratégias em condições controladas. / Customer scoring models are used to measure the credit risk of financial institution´s customers. In this work, we present three strategies that can be used to develop these models. We discuss the advantages of each of the strategies, as well as the models and statistical theory related with them. We fit models for each of these strategies using real data of a financial institution. We compare the strategies´s performance through some measures that are usually used to validate credit risk models. We still develop a simulation to study the strategies under controlled conditions.

Modelos de customer scoring são utilizados para mensurar o risco de crédito de clientes de instituições financeiras. Neste trabalho, são apresentadas três estratégias que podem ser utilizadas para o desenvolvimento desses modelos. São discutidas as vantagens de cada uma dessas estratégias, bem como os modelos e a teoria estatística associada a elas. Algumas medidas de performance usualmente utilizadas na comparação de modelos de risco de crédito são descritas. Modelos para cada uma das estratégias são ajustados utilizando-se dados reais obtidos de uma instituição financeira. A performance das estratégias para esse conjunto de dados é comparada a partir de medidas usualmente utilizadas na avaliação de modelos de risco de crédito. Uma simulação também é desenvolvida com o propósito de comparar o desempenho das estratégias em condições controladas. / Customer scoring models are used to measure the credit risk of financial institution´s customers. In this work, we present three strategies that can be used to develop these models. We discuss the advantages of each of the strategies, as well as the models and statistical theory related with them. We fit models for each of these strategies using real data of a financial institution. We compare the strategies´s performance through some measures that are usually used to validate credit risk models. We still develop a simulation to study the strategies under controlled conditions.

Sleep scoring involves classification of polysomnographic data into the various sleep stages as defined by Retschaffen and Kales. This process is time-consuming and laborious as it involves experts visually scoring the data. During recent years, there has been an increasing focus on automated sleep scoring systems and professional software programs are finding increased use. However, these systems are not relied on for scoring and are often used as a tool that facilitates easy visual scoring. This thesis proposes a neural network based approach to automatic sleep scoring using LabVIEW. Effort has been made to give the sleep expert more control over key parameters such as the frequency bands, and thus come up with scores that are more in agreement with the individual scorer than being a rigid interpretation of the R&K rules. Though this thesis is limited to the development of an offline software program, given the data acquisition facilites in LabVIEW, a complete system from data acquisition to sleep hypnograms is a fair possibility.

Thesis (MCIS - Computer and Information Sciences) -- AUT University, 2007. / Includes bibliographical references. Also held in print (x, 102 leaves : ill. ; 30 cm.) in City Campus Theses Collection (T 332.7 OET)

Thesis (Ph. D.)--Open University. BLDSC no. DX184766.

published_or_final_version / Statistics / Master / Master of Philosophy

A Thesis submitted to The University of Arizona College of Medicine - Phoenix in partial fulfillment of the requirements for the Degree of Doctor of Medicine. / Background: The International Prognostic Scoring System (IPSS) for myelodysplastic syndrome (MDS) is commonly used to predict survival and assign treatment. We explored whether markers of frailty add prognostic information to the IPSS in a cohort of older patients. Design, Setting, Participants: Retrospective cohort study of 114 MDS patients ≥ age 65 who presented to Dana‐Farber Cancer Institute between 2006‐2011 and completed a baseline quality of life questionnaire. Measurements: We evaluated questions corresponding to frailty and extracted clinical‐ pathologic data from medical records. We used Kaplan‐Meier and Cox proportional hazards models to estimate survival. Results: 114 patients consented and were available for analysis. The median age was 72.5 years, and the majority of patients were white ( 94.7%), male ( 74.6%), and over half had a Charlson comorbidity score < 2. Few patients ( 23.7%) had an IPSS score consistent with low‐risk disease and the majority received chemotherapy. In addition to traditional prognostic factors (IPSS score and history of prior chemotherapy or radiation), significant univariate predictors of survival included low serum albumin, Charlson score, the ability to take a long walk, and interference of physical symptoms in family life. The multivariate model that best predicted mortality included low serum albumin (HR=2.3; 95%CI: 1.06‐5.14), previous chemotherapy or radiation (HR=2.1; 95%CI: 1.16‐4.24), IPSS score (HR=1.7; 95%CI: 1.14‐2.49), and ease taking a long walk (HR=0.44; 95%CI: 0.23‐0.90). Conclusions: In this study of older adults with MDS, we found that markers of nutritional status and self‐reported physical function added important prognostic information to the IPSS score. More comprehensive risk assessment tools for older patients with MDS that include markers of function and frailty are needed.