Unveiling the Future Machine Learning Predicts Credit Card Scores

Abstract

The review of credit issuance decisions has undergone significant enhancements by incorporating manual judgement and statistical analysis into the decision-making processes. As financial institution databases grow in size, this integration has remarkably improved the reliability and efficiency of credit issuance decisions. Machine learning algorithms, especially Artificial Neural Network (ANN), have played a pivotal role in assisting with credit approval decisions. However, the varying algorithms and parameter selections among prediction models have led to differences in prediction performance. This study aims to improve model construction in the credit scoring process and analyze the forecast effectiveness of prevalent models. By setting a predetermined performance objective, numerous regression models and classifiers, including Decision Trees and Random Forest, were evaluated for their prediction accuracy. Through rigorous experimentation, ANN emerged as the top-performing model, exhibiting the highest performance score in terms of balanced accuracy. The findings of this research contribute to refining credit approval decision-making and offer valuable insights for financial institutions seeking to adopt robust machine learning models for credit scoring, ultimately enhancing the overall credit assessment process.