Predictive Modeling for Credit Card Fraud Detection using Supervised Algorithms

With the increase in online shopping and other advanced forms of financial fraud, detection of fraud in credit card became a major issue in the modern banking world. The new predictive model being proposed in this research combines gradient boosting with anomaly detection methods to improve performance better than the Ensemble Gradient Boosted Anomaly Detector. In dealing with intricate class imbalance transaction patterns, SMOTE employs complex feature engineering. With a score of 0.91 F1 and 0.96 AUC-ROC, which represents fantastic accuracy and recall, EGBAD outperformed the baseline technique when applied to a real dataset. With a strong architecture which prevents the system from reverting to old fraud strategies but lets it adapt to new fraud strategies; it reduces false positives to a minimum. The research approach has scalable, efficient, and accurate methodology, offering new avenues for securing the digital financial ecosystem against fraudulent attacks and therefore a tremendous contribution to fraud detection.