Modelling and Analysis of Creditworthiness of Individuals

This study is devoted to the construction of an optimal model for assessing and analyzing the creditworthiness of individuals using machine learning methods. The introduction substantiates the relevance of the research, formulates the purpose of the work and lists the tasks to be solved. In the first chapter of the work the literature on the topic of research is studied, the basic concepts of creditworthiness, as well as factors affecting creditworthiness are defined. Also, the analysis of existing approaches and modeling methods for assessing creditworthiness was carried out and the best - methods based on machine learning, the principles of which were mathematically described in the second chapter, as well as metrics for their evaluation were highlighted. In the third chapter, the initial data set was collected and analyzed and preprocessed which includes analysis for missing values, outliers, correlation analysis. Next, the selected machine learning models namely logistic regression, decision trees, random forest and gradient bousting were implemented. To evaluate the performance for each model, quality metrics were calculated and it was found that the best approach for assessing creditworthiness is the gradient bousting method using decision trees as the base algorithms. The paper contains 40 figures, 17 tables and 31 formulas.