An innovative feature selection method for support vector machines and its test on the estimation of the credit risk of default

Abstract

Support vector machines (SVM) have been extensively used for classification problems in many areas such as gene, text and image recognition. However, SVM have been rarely used to estimate the probability of default (PD) in credit risk. In this paper, we advocate the application of SVM, rather than the popular logistic regression (LR) method, for the estimation of both corporate and retail PD. Our results indicate that most of the time SVM outperforms LR in terms of classification accuracy for the corporate and retail segments. We propose a new wrapper feature selection based on maximizing the distance of the support vectors from the separating hyperplane and apply it to identify the main PD drivers. We used three datasets to test the PD estimation, containing (1) retail obligors from Germany, (2) corporate obligors from Eastern Europe, and (3) corporate obligors from Poland. Total assets, total liabilities, and sales are identified as frequent default drivers for the corporate datasets, whereas current account status and duration of the current account are frequent default drivers for the retail dataset.

Bayesian regularized artificial neural networks for the estimation of the probability of default

“…Nonetheless, not to bias our results based on a single combination of variables, we report our results for different number of variables by changing the percentile value from 0% to 90%, see Appendix 2. This is consistent with the principle applied in Sariev and Germano (2019), where model performance is assessed comparing a model on a different set of variables. Table 2 presents eight different feed-forward neural network architectures.…”

Employing Explainable AI to Optimize the Return Target Function of a Loan Portfolio

“…To bring down the number of accepted defaulters, thus increasing the TPR, we have to reject business, but this will go hand-in-hand with an increase of the FPR: since discrimination on a high level of accuracy is increasingly difficult, we will increasingly forgo non-defaulting business and thus income. Consequently, a substantial further gain in prediction accuracy is often simply not achievable, even when employing highly sophisticated ML techniques (compare, e.g., Giudici et al, 2019 ; Sariev and Germano, 2019 ; Sariev and Germano, 2020 ). Already here we can see the fundamental issue: since accepted defaults are much more costly than forgone non-defaulting business, the optimum with regard to the accuracy of predicting the number of defaults never can be the same as the optimum with regard to predicting the highest payoff.…”

A Data-driven Explainable Case-based Reasoning Approach for Financial Risk Detection

“…Especially, under the rule of the general data protection regulation (GDPR) in Europe, decision-making based solely on automated processing is prohibited, while meaningful information about the logic involved should be carried on (Voigt and Bussche 2017). To overcome this issue in the financial field, Explainable AI (XAI) models are necessary, as they provide reasons to make decisions or enable humans to understand and trust the decisions appropriately (Sariev and Germano 2019, Barredo Arrieta et al 2020, Gramespacher and Posth 2021.…”