Credit Scoring in Context of Interpretable Machine Learning

The volume Credit scoring in context of interpretable machine learning presents a unique, and simultaneously balanced, combination of explanation of theoretical concepts and contemporary scoring practices rooted in these concepts. We assume that the reader has a working analytical knowledge in fields of finance, mathematics, applied statistics and data processing algorithms. This monograph is prepared for one of the three main target groups: 1) finance and economics students (graduate or postgraduate levels), 2) risk practitioners (e.g., risk managers, data scientists, risk modelers, regulators, consultants) and 3) independent researchers (including academics, freelancing risk specialists, and machine-learning experts).

This monograph’s primary purpose and the ambition we had while writing it, were to present the development and maintenance of the creditworthiness assessment process, emphasizing the applicability of the state of the art data analytics methods (i.e., machine learning). The modern transformation allowing big data driven decision-making, especially in credits coring, is present not only in universal banks but also in many other entities that offer solutions based on trade credit. Today’s standards and business practices require full transparency and audibility of the analytical methods used. The era of black boxes, the actions of which could not be understood and decision-making processes in which stakeholders are not able to precisely explain the impact of individual factors, has already ended in many industries and in banking itself probably never came. For this reason, in this monograph so many pages are devoted not only to the machine-learning methods themselves (supported by the Explainable Artificial Intelligence - XAI) but also to other aspects of building robust and trustworthy quantitative models supporting business processes.