Assessing Lending Risk: A Comparative Study of Machine Learning Algorithms

This paper investigates credit risk assessment for a financial firm using various machine learning models. The data was preprocessed by cleaning it and features selection. To address the imbalanced data issue, eight sampling methods such as oversampling, undersampling, and over-undersampling (a hybrid method) were used. The results show that the combination of bagging ensemble learning models with the Randomoversampler method obtained the best performance in terms of accuracy (99.93%), recall (99.94%), precision (99.92%) and F1-score (99.93%). The study highlights the importance of data preparation and sampling in enhancing the performance of machine learning models. The results have implications for financial institutions seeking to make informed decisions and reduce lending risk.