This project presents a novel approach to enhance Medicare fraud detection by addressing the issue of class imbalance using Synthetic Minority Over-sampling Technique combined with Edited Nearest Neighbors (SMOTE-ENN). Traditional fraud detection methods often struggle with imbalanced datasets, where legitimate claims vastly outnumber fraudulent ones, leading to high false negative rates. Our methodology integrates SMOTE-ENN with machine learning algorithms, including Extreme Gradient Boosting (XGBoost), Adaptive Boosting (Adaboost), Light Gradient Boosting Machine (LGBM), Decision Trees (DT), Logistic Regression (LR), and Random Forest Classifiers. The project involves comprehensive data preprocessing, application of SMOTE-ENN to balance the dataset, and training of the models. Performance is evaluated using metrics such as accuracy, precision, recall, F1-score, and AUC-ROC. Experimental results demonstrate that the proposed approach significantly improves the models' ability to detect fraudulent claims, with Decision Trees achieving the highest performance. This study highlights the importance of addressing class imbalance in healthcare fraud detection and provides a robust framework for enhancing detection accuracy.

Keywords: Medicare fraud detection, Machine learning, Class imbalance, SMOTE-ENN, Synthetic Minority Over-sampling Technique.