Enhancing Credit Risk Management with Machine Learning: A Comparative Study of Predictive Models for Credit Default Prediction

Аннотация

This study investigates the application of machine learning algorithms for predictive analytics in credit risk management, aiming to enhance the accuracy of predicting credit defaults. The research compares multiple machine learning models, including logistic regression, decision trees, random forests, gradient boosting, XGBoost, and LightGBM, using a real-world credit risk dataset. The study focuses on evaluating the models' performance based on metrics such as accuracy, precision, recall, and F1-score. The results show that ensemble models, particularly XGBoost and LightGBM, outperform traditional algorithms in terms of predictive accuracy and computational efficiency, demonstrating their ability to effectively handle complex datasets. The comparative analysis highlights the strengths and weaknesses of each model, providing insights into the trade-offs between interpretability and predictive power. XGBoost and LightGBM are found to be highly effective for credit risk prediction, though The American Journal of Applied Sciences 17 https://www.theamericanjournals.com/index.php/tajas The American Journal of Applied Sciences LITERATURE REVIEW challenges such as model interpretability and overfitting remain. The findings suggest that machine The application of machine learning in credit risk learning offers a promising approach for improving management has been a subject of growing interest credit risk management, with implications for the over the past few decades. Early research focused on financial industry to make more informed, data-driven the traditional statistical methods such as logistic lending decisions. The study underscores the regression (Altman, 1968) and discriminant analysis importance of addressing interpretability concerns and (Ohlson, 1980), which laid the foundation for the field data quality issues in real-world applications, paving of credit scoring. These models used a limited set of the way for future advancements in machine learning financial ratios and historical data to predict the for credit risk prediction. likelihood of default. However, these models often struggled to capture complex relationships between Keywords: machine learning, credit risk management, variables and faced challenges in handling large and predictive analytics, XGBoost, LightGBM, decision unstructured datasets (Zhao, 2018). trees, logistic regression, model evaluation, accuracy, predictive power, data preprocessing, feature With the advent of machine learning, the landscape of selection, overfitting, interpretability. credit risk management began to shift. Machine learning models, such as decision trees (Breiman, 1986) Introduction: In recent years, credit risk management and random forests (Breiman, 2001), provided a more has become an essential aspect of financial institutions flexible and scalable alternative. Decision trees as they strive to mitigate the risks associated with modelled data through a hierarchical structure, where lending. Traditional methods of assessing credit risk each node represents a decision based on a feature, primarily rely on expert knowledge and historical and the branches represent possible outcomes. financial data. However, these methods are often Random forests, an ensemble method, combined insufficient in handling complex and large datasets. multiple decision trees to improve accuracy and reduce With the rapid advancement of machine learning overfitting. These models quickly gained popularity in techniques, financial institutions are now leveraging credit risk modeling due to their ability to handle large these technologies to improve the accuracy and datasets and capture non-linear relationships between efficiency of credit risk assessments. Predictive variables. analytics, powered by machine learning, can identify potential credit defaults more effectively by analyzing Gradient boosting, another ensemble technique, was large volumes of structured and unstructured data, introduced to further improve predictive performance. providing deeper insights into borrower behavior, and It builds a series of weak learners, where each model detecting patterns that may not be evident through corrects the errors of the previous one, allowing for conventional methods. high levels of accuracy and robustness (Friedman, 2001). This technique, implemented in models like Machine learning algorithms such as logistic regression, XGBoost (Chen & Guestrin, 2016) and LightGBM (Ke et decision trees, random forests, gradient boosting, al., 2017), has become one of the most effective support vector machines, XGBoost, and LightGBM have approaches for credit risk prediction. XGBoost, in shown remarkable success in various domains, particular, is known for its speed, scalability, and ability including credit risk modeling. These algorithms can to handle missing data and imbalanced classes, making learn from historical data, automatically adapt to new it ideal for financial applications. patterns, and make d