| Summary: | The assessment of financial credit risk constitutes an important, yet challenging research topic across multiple disciplines. This paper evaluates the risk of customers not being able to repay their obligation on time by utilizing a variety of both parametric and non-parametric (supervised) machine learning models. These methods include Decision Tree, Random Forest, Ada Boost, XG Boost, and Support Vector Machine. In addition, as a benchmark classifier, the traditional credit-risk method, Logistic Regression, was used to perform a comparison. Random Forest and XG Boost outperformed the other methods constantly, provided that thorough data analysis, pre-processing, and model-training are performed.
|
|---|