Data Mining for Predicting Creditworthiness in Credit Card Approval: A Systematic Literature Review

Abstract

The growing volume of credit card applications has led financial institutions to seek faster and more reliable methods in the approval process. Manual evaluation is not only time-consuming but also susceptible to human error, which can result in poor credit decisions and measurable financial losses. This study conducts a Systematic Literature Review (SLR) to examine data mining techniques applied to creditworthiness prediction. Five research questions were formulated to identify: (1) commonly used data mining techniques, (2) frequently used datasets, (3) performance evaluation metrics, (4) algorithms with the strongest performance, and (5) recurring challenges and practical recommendations. A structured search across three academic databases — Scopus, Google Scholar, and GARUDA — yielded 8 relevant articles (7 primary experimental studies and 1 secondary study) published between 2021 and 2025. The findings show that Naïve Bayes, Decision Tree, Random Forest, Support Vector Machine, and K-Nearest Neighbors are the most widely applied methods. Tree-based algorithms such as Decision Tree and Random Forest consistently yield high accuracy, while K-Nearest Neighbors also delivers strong results in specific experimental settings. Naïve Bayes appears most frequently across studies, and its performance can be improved through metaheuristic approaches such as Particle Swarm Optimization (PSO). Standard evaluation metrics include accuracy, precision, recall, F1-score, and AUC-ROC. The review underscores the importance of data preprocessing, class imbalance handling, and hyperparameter tuning in building reliable prediction models — findings with direct implications for financial institutions seeking to reduce non-performing loan rates.