Bank's Loan Portfolio Analysis Using Machine Learning Techniques

Big Data technologies, including machine learning algorithms, are widespread in the financial sector. Machine learning in both Russian and international banks helps to solve problems from evaluating credit ratings and tracking fraudulent transactions to replacing personal financial advisors with bots. The processes of assessing credit risk and evaluating borrowers are well studied and automated. However, due to the high competition in the financial market, many banks seek to improve valuation approaches using the accumulated data, which include loan portfolio. The main goal of this work is to develop and compare models for predicting the solvency of a borrower based on the influence of specific features of client data, as well as determining a set of critical features that have the maximum correlation with the probability of loan repayment. Using the collected features, it is possible to more quickly conduct an initial customer assessment when considering a loan application or generating a marketing offer to existing clients.