Complex Model of Solvency Assesment of a Bank Customer

In this thesis, work was carried out on a comprehensive assessment of the solvency of a client of a Russian commercial bank. Work was carried out to clean, process and analyze data received from three credit bureaus at the request of the bank. Based on the received data, including strings with financial indicators and contract features as well as rows with aggregated data on the payment history, factors influencing the client's reliability are formed. Based on the factors and characteristics of each individual person, models of logistic regression and decision trees are constructed. A comparison of the results obtained and the features of the models is made, the actions performed being called the complex method. A set of models is proposed and an analysis of the predictive accuracy/power of the method is carried out. The model obtained differs significantly from such models as the Altman model and the Chesser model. Keywords. Logistic regression model, decision tree, solvency prediction, data sampling.