A new advanced fraud detection system for credit card transactions is developed through the combination of K-SMOTEENN and a Stacking Ensemble model according to the paper. This method uses K-SMOTEENN to detect minority class effectively while the Stacking Ensemble brings together Random Forest and Decision Trees and XGBoost and CNN and LSTM to deliver powerful classification results. The fraud detection accuracy benefits from CNN and LSTM which extract spatio-temporal patterns through the model features. The proposed system outranks traditional techniques while offering Explainable AI (XAI) interpretation of the model because it performs better regarding precision, recall, and F1-score metrics. The methodology succeeds in detecting fraudulent financial transactions through its experimental results.