Banking business is associated with risks, which is unavoidable. In the simplest terms, risk is the uncertainty of an event that may occur in the future and in the case of a bank, the uncertainty of the outcome, of business investments. Different types of banking risks can be classified as strategic risk, compliance risk, credit risk, cyber security risk, liquidity risk, market risk, operational risk, etc. Among these, credit risk represents the most important type of risk for commercial banks. Credit risk is simply understood as the risk that a bank assumes when lending money to borrowers. Borrowers may default on their debts, causing a loss to the bank. Credit modeling is the process of preparing a statistical model for predicting the probability of non-payment of liabilities based on historical data.
In my masters thesis, I present three different approaches to the development of a model for predicting the probability of default for retail banking services. Banking services for the population consist of postal credit, housing credit and credit cards. Two of the models are based on logistic regression, which is the most popular family of models in the field of predicting the probability of default. The third approach represents building a model based on a random forest machine learning algorithm. We have described all approaches and presented their final predictive metrics. The models were prepared and tested on the historical data of Nova Ljubljanska banka d.d. We implemented the algorithms in the Python programming language based on the Scikit-learn library.
|