Year of Graduation
Credit Scoring Problem Solution via Modified Machine Learning Models Based on Data of Legal Entities - Borrowers of Top Tier Russian Banks
This paper explores various machine learning methods for credit scoring problem. An important part of the paper devoted to economic sense of small-scale improvements in model quality. The data used for measuring the effectiveness of different algorithms consists of largest Russian banks’ observations of small business clients. Also this paper covers the analysis of an algorithm based on modified cost function which delivers the rank-ordering performance of the same kind as uninterpretable black-box algorithms. The paper covers relevant literature and studies review.