Senzi Mofokeng

Abstract
The objective of this study is to evaluate the transparency of
the credit verification process when machine learning
algorithms are used to predict customer credit facility defaults.
XGBoost, was utilised for enhancing credit score evaluation on
secondary credit verification data obtained from Kaggle.
Meanwhile, the Local Interpretable Model-Agnostic
Explanation (LIME) provides valuable insights into model
operations, enabling the identification of critical areas within
images or highlighting important features. The results indicate
that the most important feature is the duration, also known as
the term of the loan. The second important feature is the
paydays, which is the number of days in which repayments are
made, and the third most important feature is whether the
customer owns a house.