| Abstract [eng] |
This master’s final project examines credit risk assessment methods for the “EstateGuru“ crowdfunding platform, focusing on the segment of real estate loans. The relevance of the study is driven by the rapid growth of the peer-to-peer (P2P) lending market and the increasing need for accurate default risk evaluation in order to protect investors’ capital and optimize investment decisions. The aim of the paper is to develop and compare credit risk assessment models for “EstateGuru“ platform data by integrating both loan-specific characteristics and macroeconomic indicators, as well as to evaluate their applicability for investment strategy formation. During the research, loan-level data from the “EstateGuru“ platform were collected and prepared including data cleaning, transformation, and feature engineering procedures. Macroeconomic indicators were additionally integrated into the models, including inflation, interest rates, gross domestic product, and changes in real estate price indexes. Credit risk assessment was performed using logistic regression, Support Vector Machine (SVM), Random Forest, and CatBoost models. Model performance was evaluated using detection curves (ROC, DET, and Precision-Recall curves) together with various classification accuracy metrics. The results of the study showed that the Random Forest model achieved the best predictive performance and most effectively identified high-risk loans. Variable importance analysis revealed that property value, refinancing indicator, loan-to-value ratio (LTV), country indicator, and several macroeconomic variables had the greatest influence on model predictions. Additional SHAP analysis provided insights into the direction of variable effects and showed that higher property values were associated with a lower probability of loan default. Based on the developed credit risk models, investment strategies were constructed to filter higher-risk loans and optimize the risk-return trade-off. The findings suggest that the integration of advanced machine learning methods and macroeconomic indicators can improve the quality of credit risk assessment in the P2P real estate financing market. |
