{"title":"Deep Learning Based Measure of Name Concentration Risk","authors":"Eva Lütkebohmert, Julian Sester","doi":"arxiv-2403.16525","DOIUrl":"https://doi.org/arxiv-2403.16525","url":null,"abstract":"We propose a new deep learning approach for the quantification of name\u0000concentration risk in loan portfolios. Our approach is tailored for small\u0000portfolios and allows for both an actuarial as well as a mark-to-market\u0000definition of loss. The training of our neural network relies on Monte Carlo\u0000simulations with importance sampling which we explicitly formulate for the\u0000CreditRisk${+}$ and the ratings-based CreditMetrics model. Numerical results\u0000based on simulated as well as real data demonstrate the accuracy of our new\u0000approach and its superior performance compared to existing analytical methods\u0000for assessing name concentration risk in small and concentrated portfolios.","PeriodicalId":501128,"journal":{"name":"arXiv - QuantFin - Risk Management","volume":"52 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-03-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140298803","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Risk exchange under infinite-mean Pareto models","authors":"Yuyu Chen, Paul Embrechts, Ruodu Wang","doi":"arxiv-2403.20171","DOIUrl":"https://doi.org/arxiv-2403.20171","url":null,"abstract":"We study the optimal decisions of agents who aim to minimize their risks by\u0000allocating their positions over extremely heavy-tailed (i.e., infinite-mean)\u0000and possibly dependent losses. The loss distributions of our focus are\u0000super-Pareto distributions which include the class of extremely heavy-tailed\u0000Pareto distributions. For a portfolio of super-Pareto losses,\u0000non-diversification is preferred by decision makers equipped with well-defined\u0000and monotone risk measures. The phenomenon that diversification is not\u0000beneficial in the presence of super-Pareto losses is further illustrated by an\u0000equilibrium analysis in a risk exchange market. First, agents with super-Pareto\u0000losses will not share risks in a market equilibrium. Second, transferring\u0000losses from agents bearing super-Pareto losses to external parties without any\u0000losses may arrive at an equilibrium which benefits every party involved. The\u0000empirical studies show that extremely heavy tails exist in real datasets.","PeriodicalId":501128,"journal":{"name":"arXiv - QuantFin - Risk Management","volume":"124 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-03-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140602316","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Spanning Multi-Asset Payoffs With ReLUs","authors":"Sébastien BossuLPSM, UPCité, Stéphane CrépeyLPSM, UPCité, Hoang-Dung NguyenLPSM, UPCité","doi":"arxiv-2403.14231","DOIUrl":"https://doi.org/arxiv-2403.14231","url":null,"abstract":"We propose a distributional formulation of the spanning problem of a\u0000multi-asset payoff by vanilla basket options. This problem is shown to have a\u0000unique solution if and only if the payoff function is even and absolutely\u0000homogeneous, and we establish a Fourier-based formula to calculate the\u0000solution. Financial payoffs are typically piecewise linear, resulting in a\u0000solution that may be derived explicitly, yet may also be hard to numerically\u0000exploit. One-hidden-layer feedforward neural networks instead provide a natural\u0000and efficient numerical alternative for discrete spanning. We test this\u0000approach for a selection of archetypal payoffs and obtain better hedging\u0000results with vanilla basket options compared to industry-favored approaches\u0000based on single-asset vanilla hedges.","PeriodicalId":501128,"journal":{"name":"arXiv - QuantFin - Risk Management","volume":"31 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-03-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140203703","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"A Markov approach to credit rating migration conditional on economic states","authors":"Michael Kalkbrener, Natalie Packham","doi":"arxiv-2403.14868","DOIUrl":"https://doi.org/arxiv-2403.14868","url":null,"abstract":"We develop a model for credit rating migration that accounts for the impact\u0000of economic state fluctuations on default probabilities. The joint process for\u0000the economic state and the rating is modelled as a time-homogeneous Markov\u0000chain. While the rating process itself possesses the Markov property only under\u0000restrictive conditions, methods from Markov theory can be used to derive the\u0000rating process' asymptotic behaviour. We use the mathematical framework to\u0000formalise and analyse different rating philosophies, such as point-in-time\u0000(PIT) and through-the-cycle (TTC) ratings. Furthermore, we introduce stochastic\u0000orders on the bivariate process' transition matrix to establish a consistent\u0000notion of \"better\" and \"worse\" ratings. Finally, the construction of PIT and\u0000TTC ratings is illustrated on a Merton-type firm-value process.","PeriodicalId":501128,"journal":{"name":"arXiv - QuantFin - Risk Management","volume":"59 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-03-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140298771","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Uncertainty in the financial market and application to forecastabnormal financial fluctuations","authors":"Shige Peng, Shuzhen Yang, Wenqing Zhang","doi":"arxiv-2403.12647","DOIUrl":"https://doi.org/arxiv-2403.12647","url":null,"abstract":"The integration and innovation of finance and technology have gradually\u0000transformed the financial system into a complex one. Analyses of the causesd of\u0000abnormal fluctuations in the financial market to extract early warning\u0000indicators revealed that most early warning systems are qualitative and causal.\u0000However, these models cannot be used to forecast the risk of the financial\u0000market benchmark. Therefore, from a quantitative analysis perspective, we focus\u0000on the mean and volatility uncertainties of the stock index (benchmark) and\u0000then construct three early warning indicators: mean uncertainty, volatility\u0000uncertainty, and ALM-G-value at risk. Based on the novel warning indicators, we\u0000establish a new abnormal fluctuations warning model, which will provide a\u0000short-term warning for the country, society, and individuals to reflect in\u0000advance.","PeriodicalId":501128,"journal":{"name":"arXiv - QuantFin - Risk Management","volume":"24 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-03-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140169281","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Giuseppe Calafiore, Giulia Fracastoro, Anton Proskurnikov
{"title":"Default Resilience and Worst-Case Effects in Financial Networks","authors":"Giuseppe Calafiore, Giulia Fracastoro, Anton Proskurnikov","doi":"arxiv-2403.10631","DOIUrl":"https://doi.org/arxiv-2403.10631","url":null,"abstract":"In this paper we analyze the resilience of a network of banks to joint price\u0000fluctuations of the external assets in which they have shared exposures, and\u0000evaluate the worst-case effects of the possible default contagion. Indeed, when\u0000the prices of certain external assets either decrease or increase, all banks\u0000exposed to them experience varying degrees of simultaneous shocks to their\u0000balance sheets. These coordinated and structured shocks have the potential to\u0000exacerbate the likelihood of defaults. In this context, we introduce first a\u0000concept of {default resilience margin}, $epsilon^*$, i.e., the maximum\u0000amplitude of asset prices fluctuations that the network can tolerate without\u0000generating defaults. Such threshold value is computed by considering two\u0000different measures of price fluctuations, one based on the maximum individual\u0000variation of each asset, and the other based on the sum of all the asset's\u0000absolute variations. For any price perturbation having amplitude no larger than\u0000$epsilon^*$, the network absorbs the shocks remaining default free. When the\u0000perturbation amplitude goes beyond $epsilon^*$, however, defaults may occur.\u0000In this case we find the worst-case systemic loss, that is, the total unpaid\u0000debt under the most severe price variation of given magnitude. Computation of\u0000both the threshold level $epsilon^*$ and of the worst-case loss and of a\u0000corresponding worst-case asset price scenario, amounts to solving suitable\u0000linear programming problems.}","PeriodicalId":501128,"journal":{"name":"arXiv - QuantFin - Risk Management","volume":"30 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-03-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140169145","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Improving Fairness in Credit Lending Models using Subgroup Threshold Optimization","authors":"Cecilia Ying, Stephen Thomas","doi":"arxiv-2403.10652","DOIUrl":"https://doi.org/arxiv-2403.10652","url":null,"abstract":"In an effort to improve the accuracy of credit lending decisions, many\u0000financial intuitions are now using predictions from machine learning models.\u0000While such predictions enjoy many advantages, recent research has shown that\u0000the predictions have the potential to be biased and unfair towards certain\u0000subgroups of the population. To combat this, several techniques have been\u0000introduced to help remove the bias and improve the overall fairness of the\u0000predictions. We introduce a new fairness technique, called textit{Subgroup\u0000Threshold Optimizer} (textit{STO}), that does not require any alternations to\u0000the input training data nor does it require any changes to the underlying\u0000machine learning algorithm, and thus can be used with any existing machine\u0000learning pipeline. STO works by optimizing the classification thresholds for\u0000individual subgroups in order to minimize the overall discrimination score\u0000between them. Our experiments on a real-world credit lending dataset show that\u0000STO can reduce gender discrimination by over 90%.","PeriodicalId":501128,"journal":{"name":"arXiv - QuantFin - Risk Management","volume":"4 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-03-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140169584","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Daixin Wang, Zhiqiang Zhang, Yeyu Zhao, Kai Huang, Yulin Kang, Jun Zhou
{"title":"Financial Default Prediction via Motif-preserving Graph Neural Network with Curriculum Learning","authors":"Daixin Wang, Zhiqiang Zhang, Yeyu Zhao, Kai Huang, Yulin Kang, Jun Zhou","doi":"arxiv-2403.06482","DOIUrl":"https://doi.org/arxiv-2403.06482","url":null,"abstract":"User financial default prediction plays a critical role in credit risk\u0000forecasting and management. It aims at predicting the probability that the user\u0000will fail to make the repayments in the future. Previous methods mainly extract\u0000a set of user individual features regarding his own profiles and behaviors and\u0000build a binary-classification model to make default predictions. However, these\u0000methods cannot get satisfied results, especially for users with limited\u0000information. Although recent efforts suggest that default prediction can be\u0000improved by social relations, they fail to capture the higher-order topology\u0000structure at the level of small subgraph patterns. In this paper, we fill in\u0000this gap by proposing a motif-preserving Graph Neural Network with curriculum\u0000learning (MotifGNN) to jointly learn the lower-order structures from the\u0000original graph and higherorder structures from multi-view motif-based graphs\u0000for financial default prediction. Specifically, to solve the problem of weak\u0000connectivity in motif-based graphs, we design the motif-based gating mechanism.\u0000It utilizes the information learned from the original graph with good\u0000connectivity to strengthen the learning of the higher-order structure. And\u0000considering that the motif patterns of different samples are highly unbalanced,\u0000we propose a curriculum learning mechanism on the whole learning process to\u0000more focus on the samples with uncommon motif distributions. Extensive\u0000experiments on one public dataset and two industrial datasets all demonstrate\u0000the effectiveness of our proposed method.","PeriodicalId":501128,"journal":{"name":"arXiv - QuantFin - Risk Management","volume":"68-69 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-03-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140107270","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"On Geometrically Convex Risk Measures","authors":"Mücahit Aygün, Fabio Bellini, Roger J. A. Laeven","doi":"arxiv-2403.06188","DOIUrl":"https://doi.org/arxiv-2403.06188","url":null,"abstract":"Geometrically convex functions constitute an interesting class of functions\u0000obtained by replacing the arithmetic mean with the geometric mean in the\u0000definition of convexity. As recently suggested, geometric convexity may be a\u0000sensible property for financial risk measures ([7,13,4]). We introduce a notion of GG-convex conjugate, parallel to the classical\u0000notion of convex conjugate introduced by Fenchel, and we discuss its\u0000properties. We show how GG-convex conjugation can be axiomatized in the spirit\u0000of the notion of general duality transforms introduced in [2,3]. We then move to the study of GG-convex risk measures, which are defined as\u0000GG-convex functionals defined on suitable spaces of random variables. We derive\u0000a general dual representation that extends analogous expressions presented in\u0000[4] under the additional assumptions of monotonicity and positive homogeneity.\u0000As a prominent example, we study the family of Orlicz risk measures. Finally,\u0000we introduce multiplicative versions of the convex and of the increasing convex\u0000order and discuss related consistency properties of law-invariant GG-convex\u0000risk measures.","PeriodicalId":501128,"journal":{"name":"arXiv - QuantFin - Risk Management","volume":"34 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-03-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140107112","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Rambod Rahmani, Marco Parola, Mario G. C. A. Cimino
{"title":"A machine learning workflow to address credit default prediction","authors":"Rambod Rahmani, Marco Parola, Mario G. C. A. Cimino","doi":"arxiv-2403.03785","DOIUrl":"https://doi.org/arxiv-2403.03785","url":null,"abstract":"Due to the recent increase in interest in Financial Technology (FinTech),\u0000applications like credit default prediction (CDP) are gaining significant\u0000industrial and academic attention. In this regard, CDP plays a crucial role in\u0000assessing the creditworthiness of individuals and businesses, enabling lenders\u0000to make informed decisions regarding loan approvals and risk management. In\u0000this paper, we propose a workflow-based approach to improve CDP, which refers\u0000to the task of assessing the probability that a borrower will default on his or\u0000her credit obligations. The workflow consists of multiple steps, each designed\u0000to leverage the strengths of different techniques featured in machine learning\u0000pipelines and, thus best solve the CDP task. We employ a comprehensive and\u0000systematic approach starting with data preprocessing using Weight of Evidence\u0000encoding, a technique that ensures in a single-shot data scaling by removing\u0000outliers, handling missing values, and making data uniform for models working\u0000with different data types. Next, we train several families of learning models,\u0000introducing ensemble techniques to build more robust models and hyperparameter\u0000optimization via multi-objective genetic algorithms to consider both predictive\u0000accuracy and financial aspects. Our research aims at contributing to the\u0000FinTech industry in providing a tool to move toward more accurate and reliable\u0000credit risk assessment, benefiting both lenders and borrowers.","PeriodicalId":501128,"journal":{"name":"arXiv - QuantFin - Risk Management","volume":"24 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-03-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140056777","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}