29 October 2020
Scoring models are widely used to support resource allocation decisions in finance. Credit scorecards are a prominent example. They are trained on the data of previously accepted applicants, whose repayment behaviour has been observed. This creates sampling bias: the training data represent a limited region of the distribution of borrowers to which the model is applied in operations.
The paper contributes to the field of reject inference as follows. First, we quantify loss due to the sampling bias through a simulation study and demonstrate how the bias affects the training and the evaluation of scorecards. Second, we propose a shallow self-learning framework that tackles the bias during training and mitigates the performance loss by inferring labels of selected rejected applicants. Last, we design a new evaluation metric denoted as kickout measure. This measure reduces the bias between evaluating a scorecard on a biased training sample of accepted clients and a test sample representative of the operating conditions of the scorecard. Experiments on real credit scoring data confirm the superiority of our propositions over previous bias correction methods in terms of predictive performance and profitability.
Stefan received a diploma in business administration (2002) and a PhD (2007) from the University of Hamburg. He worked as a senior lecturer in business informatics at the Institute of Information Systems of the University of Hamburg. Since 2008, Stefan has been a guest lecturer at the School of Management of the University of Southampton, where he gives undergraduate and postgraduate courses on quantitative methods, electronic business, and web application development.
Stefan completed his habilitation on decision analysis and support using ensemble forecasting models in 2012. He then joined the Humboldt-University of Berlin in 2014, where he heads the Chair of Information Systems at the School of Business and Economics.
He serves as associate editor for the International Journal of Business Analytics, Digital Finance, and the International Journal of Forecasting, as well as department editor of Business and Information System Engineering (BISE). Stefan has secured substantial amounts of research funding and published several papers in leading international journals and conferences, including the European Journal of Operational Research, the IEEE Transactions of Software Engineering, and the International Conference on Information Systems.
Stefan’s research concerns the support of managerial decision-making using quantitative empirical methods. He specialises in applications of (deep) machine learning techniques in the broad scope of marketing and risk analytics. He actively participates in knowledge transfer and consulting projects with industry partners; from startup companies to global players and not-for-profit organisations.