595 research outputs found

    Inteligência de decisão no processo de avaliação de qualificação de crédito

    No full text
    For a company considering automating its decisions with machine learning, there are two fundamental considerations. First, the associated risks and rewards of this technology need to be evaluated. Second, as businesses often refrain from adopting machine learning in their processes due to lack of understanding, the issue of the interpretability of ML models must be addressed. The aim of the thesis is to automate credit qualification assessment for increased decision-making speed and quality. To do that, we summarize the literature regarding the use of ML in credit qualification. We then build a prediction model by combining state-of-the-art Bagging and Boosting algorithms including XGBoost and LightGBM and demonstrate their performance on a large dataset from a German bank. The methodology consists of data preparation, model construction and state-of-the-art hyperparameter optimization. These models are then combined into a heterogenous ensemble through a Bagging algorithm. For performance evaluation, we use several state-of-the-art metrics. We find LightGBM to be the best performing individual classifier and the heterogenous ensemble to outperform the individual models. While the out-of-sample performance of the models show a noticeable decrease, we conclude that partial automation of the credit qualification assessment based on the model is feasible and profitable. To address the explainability of the Classifier’s decisions, state-of-the-art model interpretation software SHAP is utilized. We show that the overall model as well as individual decisions can be interpreted, also shedding light on implicit decision factors that are present in the current manual process.As empresas que automatizam as suas decisões com a aprendizagem automática têm de fazer duas considerações fundamentais. Em primeiro lugar, é necessário avaliar os riscos e os benefícios associados a esta tecnologia. Em segundo lugar, como as empresas muitas vezes se abstêm de integrar a aprendizagem automática nos seus processos devido à falta de compreensão, é necessário abordar a questão dos modelos de aprendizagem automática interpretáveis. O objetivo desta tese é automatizar a avaliação da qualificação do crédito para aumentar a velocidade e a qualidade da tomada de decisões. Para isso, é feita uma síntese da literatura sobre o uso de ML na qualificação de crédito. Em seguida, constrói-se um modelo de previsão com algoritmos de Bagging e Boosting, incluindo XGBoost e LightGBM, e demonstra-se o seu desempenho num grande conjunto de dados bancários. A metodologia consiste na preparação de dados, construção de modelos e otimização de hiperparâmetros. Estes modelos são depois combinados num conjunto heterogéneo através de um algoritmo de Bagging. Para a avaliação do desempenho, utilizamos várias métricas. Verificamos que o LightGBM é o melhor classificador e que o conjunto heterogéneo por vezes supera os modelos individuais. Embora o desempenho fora da amostra dos modelos mostre uma diminuição, concluímos que a automatização da avaliação da qualificação do crédito com base no modelo é viável. Em última análise, é utilizado o software de interpretação de modelos SHAP, que é o mais avançado. Mostramos que o modelo global e as decisões individuais podem ser interpretados, revelando factores de decisão implícitos

    The Effects of Avatar and Environment on Thermal Perception and Skin Temperature in Virtual Reality

    No full text
    Humans’ thermal regulation and subjective perception of temperature is highly plastic and depends on the visual appearance of the surrounding environment. Previous work shows that an environment’s color temperature affects the experienced temperature. As virtual reality (VR) enables visual immersion, recent work suggests that a VR scene’s color temperature also affects experienced temperature. It is, however, unclear if an avatar’s appearance also affects users’ thermal perception and if a change in thermal perception even influences the body temperature. Therefore, we conducted a study with 32 participants performing a task in an ice or fire world while having ice or fire hands. We show that being in a fire world or having fire hands increases the perceived temperature. We even show that having fire hands decreases the hand temperature compared to having ice hands. We discuss the implications for the design of VR systems and future research directions

    Of pits and ditches

    No full text
    In dieser Diplomarbeit werden Bewegung und Materialtransformation als miteinander verbundene künstlerische Prozesse untersucht. Im Zentrum steht ein Feldexperiment, welches die Entstehung von Einzelelementen zur Sammlung in einer Reihe von Exkursionen begleitet. Dabei wurden aufgelassene Abbaustätten erkundet und Materialien gesammelt, welche während des Transports durch eine ergebnisoffene Methode verändert wurden. Die Formgebung der Materialien ergibt sich dabei durch ihre eigenen geologischen und physikalischen Eigenschaften sowie die Kräfte der Bewegung. Die für den Transport notwendige Energie stammt vom Verfasser selbst, der die Materialien mit dem Fahrrad transportiert hat. Die entstandene Sammlung macht die zurückgelegten Wege in unmittelbarer Form sichtbar. Abschließend reflektiert die Arbeit den Prozess anhand eines Interviews und der praktischen Erfahrungen während des Feldexperiments.This thesis examines movement and material transformation as interconnected artistic processes. It focuses on a field experiment that accompanies the creation of individual elements for collection in a series of excursions. Abandoned mining sites were explored and materials collected, which were then altered during transport using an open-ended method. The shape of the materials is determined by their own geological and physical properties as well as the forces of movement. The energy required for transport came from the author himself, who transported the materials by bicycle. The resulting collection makes the distances traveled visible in an immediate form. Finally, the thesis reflects on the process based on an interview and practical experiences during the field experiment

    Essays in applied microeconomics. Research collaboration and dissemination in economics

    No full text
    This dissertation studies the production and dissemination of economic research in three essays. The chapters explore data on economic research articles and their authors to study, first, the effect of national borders on the international dissemination of economic research along cultural and technological links, second, the rise of collaborations between economists, and, third, the benefit of local research clusters. Chapter 1 provides an introduction and outlines the specific research questions that are explored in this dissertation. Chapter 2 examines the existence of a border effect in the dissemination of knowledge in economics using a gravity model framework and novel data on domestic and international citation flows in economic research between 1970 and 2016. The citation data are linked to geographic and cultural distance measures, as well as novel indicators for virtual proximity and English proficiency. The results show that (i) citation patterns follow the law of gravity; (ii) citations in economics exhibit a strong and significant home bias by an overall factor of 1.9 for all ten leading countries (a 90 percent higher propensity to cite domestic articles); (iii) bilaterally low levels of English proficiency are associated with a cost of up to 30 percent, while similarity in English proficiency is insignificant for the total sample; (iv) countries with closer early internet ties have significantly higher shares of bilateral citations of up to .25 percent for a 1 percent increase in internet hyperlinks. Over time, the estimated impact of home bias as well as geographic and cultural distance declines but remains significant. Chapter 3 presents new evidence on several features of co-authorship in articles in economic journals. It builds on previous work by reviewing the key literature relating to the rise in co-authorship in economics. The empirical analysis draws on around 175,000 articles in the top 255 journals, over the period 1996 to 2014. The rises in quarto-plus and cross-country co-authored papers are striking, as are the differences in citations per article and citations per author. There is evidence of an alphabetical ordering of authors as the standard in co-authored papers in top journals with no downward trend evident over time. A correlation between co-authorship and career stage is observed with young authors publishing significantly more solo-authored articles. Chapter 4 studies benefits from local research linkages. Recent research observes a declining or no return from the quality of local colleagues on a researcher\u27s productivity but positive spill-overs between co-authors. The findings of this chapter show benefits from research linkages between department colleagues below the level of co-authorship. Using data from the CVs of around 1,000 highly-cited economists, this chapter tests whether take-up of a research article by future research is increased if it links with the research of local colleagues. The estimates show that ``high-profile\u27\u27 research that references articles by local colleagues receives significantly more citations than comparable work by the same authors. There is no dissemination benefit for less prominently published ``routine-type\u27\u27 research or for colleagues without a thematic link. Chapter 5 concludes the dissertation

    Corporate governance in Germany : an economic perspective

    No full text
    A financial system can only perform its function of channelling funds from savers to investors if it offers sufficient assurance to the providers of the funds that they will reap the rewards which have been promised to them. To the extent that this assurance is not provided by contracts alone, potential financiers will want to monitor and influence managerial decisions. This is why corporate governance is an essential part of any financial system. It is almost obvious that providers of equity have a genuine interest in the functioning of corporate governance. However, corporate governance encompasses more than investor protection. Similar considerations also apply to other stakeholders who invest their resources in a firm and whose expectations of later receiving an appropriate return on their investment also depend on decisions at the level of the individual firm which would be extremely difficult to anticipate and prescribe in a set of complete contingent contracts. Lenders, especially long-term lenders, are one such group of stakeholders who may also want to play a role in corporate governance; employees, especially those with high skill levels and firm-specific knowledge, are another. The German corporate governance system is different from that of the Anglo-Saxon countries because it foresees the possibility, and even the necessity, to integrate lenders and employees in the governance of large corporations. The German corporate governance system is generally regarded as the standard example of an insider-controlled and stakeholder-oriented system. Moreover, only a few years ago it was a consistent system in the sense of being composed of complementary elements which fit together well. The first objective of this paper is to show why and in which respect these characterisations were once appropriate. However, the past decade has seen a wave of developments in the German corporate governance system, which make it worthwhile and indeed necessary to investigate whether German corporate governance has recently changed in a fundamental way. More specifically one can ask which elements and features of German corporate governance have in fact changed, why they have changed and whether those changes which did occur constitute a structural change which would have converted the old insider-controlled system into an outsider-controlled and shareholder-oriented system and/or would have deprived it of its former consistency. It is the second purpose of this paper to answer these questions

    Networks of micro and small enterprise banks : a contribution to financial sector development : [Version March 2005]

    No full text
    The paper is a follow-up to an article published in Technique Financière et Developpement in 2000 (see the appendix to the hardcopy version), which portrayed the first results of a new strategy in the field of development finance implemented in South-East Europe. This strategy consists in creating microfinance banks as greenfield investments, that is, of building up new banks which specialise in providing credit and other financial services to micro and small enterprises, instead of transforming existing credit-granting NGOs into formal banks, which had been the dominant approach in the 1990s. The present paper shows that this strategy has, in the course of the last five years, led to the emergence of a network of microfinance banks operating in several parts of the world. After discussing why financial sector development is a crucial determinant of general social and economic development and contrasting the new strategy to former approaches in the area of development finance, the paper provides information about the shareholder composition and the investment portfolio of what is at present the world's largest and most successful network of microfinance banks. This network is a good example of a well-functioning "private public partnership". The paper then provides performance figures and discusses why the creation of such a network seems to be a particularly promising approach to the creation of financially self-sustaining financial institutions with a clear developmental objective

    Networks of micro and small enterprise banks : a contribution to financial sector development :[Version January 2004]

    No full text
    The paper is a follow-up to an article published in Technique Financière et Developpement in 2000 (see the appendix to the hardcopy version), which portrayed the first results of a new strategy in the field of development finance implemented in South-East Europe. This strategy consists in creating microfinance banks as greenfield investments, that is, of building up new banks which specialise in providing credit and other financial services to micro and small enterprises, instead of transforming existing credit-granting NGOs into formal banks, which had been the dominant approach in the 1990s. The present paper shows that this strategy has, in the course of the last five years, led to the emergence of a network of microfinance banks operating in several parts of the world. After discussing why financial sector development is a crucial determinant of general social and economic development and contrasting the new strategy to former approaches in the area of development finance, the paper provides information about the shareholder composition and the investment portfolio of what is at present the world's largest and most successful network of microfinance banks. This network is a good example of a well-functioning "private public partnership". The paper then provides performance figures and discusses why the creation of such a network seems to be a particularly promising approach to the creation of financially self-sustaining financial institutions with a clear developmental objective

    Cloud top altitudes observed with airborne lidar during the ACLOUD campaign

    No full text
    During the ACLOUD (Arctic CLoud Observations Using airborne measurements during polar Day) campaign conducted in May / June 2017, the Airborne Mobile Aerosol Lidar (AMALi) was used to measure the cloud top altitude of clouds below the aircraft Polar 5. The data set provides cloud top altitude along the flight tracks of 14 flights performed by the Polar 5 aircraft in 5 Hz resolution
    corecore