<![CDATA[Loan approval is an important process in financial institutions because it relates to credit risk and applicant eligibility. This study classifies loan approval using the Random Forest algorithm on the Kaggle Loan Risk Prediction dataset and evaluates the effect of Synthetic Minority Oversampling Technique (SMOTE) on model performance. The dataset contains 5,000 records with an imbalanced class distribution: 76.98% rejected loans and 23.02% approved loans. The research stages include preprocessing, missing value imputation, one-hot encoding, an 80:20 stratified train-test split, SMOTE, Random Forest modeling, hyperparameter tuning, model evaluation, and feature importance analysis. The results show that Random Forest without SMOTE achieved the best direct classification performance with accuracy of 0.9640, precision of 0.9533, recall of 0.8870, F1-score of 0.9189, and ROC-AUC of 0.9349. SMOTE therefore needs to be evaluated according to the model objective.]]>
Copyrights © 2026
