<![CDATA[This study evaluates the performance of the Support Vector Machine (SVM) algorithm in detecting credit card fraud by overcoming the class imbalance problem using the Synthetic Minority Oversampling Technique (SMOTE) technique and parameter optimization through Grid Search. The dataset used is sourced from Kaggle, consists of 10,001 transactions, and has been balanced. SMOTE is applied exclusively to the training data to prevent data leakage. The optimization process produces the best parameters at a value of C = 10 and gamma = 0.1. Model evaluation is carried out using recall, precision, F1-score, and AUC-ROC metrics. The results show a significant increase in performance in recognizing fraudulent transactions. The final model recorded a recall of 0.68, precision 0.90, F1-score 0.77, and AUC-ROC 0.98. These findings prove that the combination of SMOTE techniques and parameter optimization can improve the effectiveness of SVM in classifying minority classes more accurately. This approach is considered to have great potential to be applied in automated fraud detection systems in the financial sector.]]>
Copyrights © 2025
