Improving Credit Score Classification Using Predictive Analysis and Machine Learning Techniques

Credit score classification is crucial to financial decision-making, allowing institutions to evaluate creditworthiness and manage risk accurately. Statistical models, on which conventional credit scoring techniques are often based, have difficulty capturing subtle patterns in credit information. This research examines the performance of various ML models such as Decision Trees, Logistic Regression, SVM, Random Forest, LightGBM, Gradient Boosting (GB), AdaBoost, XGBoost and Naive Bayes (NB) on three benchmark datasets: Australian Credit dataset, German Credit dataset, and Kaggle Credit Score Classification dataset. All the models are run through feature selection, hyperparameter tuning to achieve improved prediction accuracy. Experimental findings reveal that Random Forest performs significantly better than all other models with the highest accuracy of 90.58% on the Australian Credit dataset, and XGBoost performed next best but at a slightly lower level than RF model. The above findings are indicative of the power of ensemble learning methods in credit risk assessment and offer a solid model for data-driven financial decision-making and mitigating lending risks for banks.