Leaveraging Machine Learning Approach for Improved Medicare Fraud Detection using Advanced ML Technique

Abstract

Medicare fraud poses serious challenges by causing notable financial losses and weakening the trust in healthcare systems. Traditional detection approaches often fall short due to the sophisticated and constantly evolving nature of fraudulent schemes. This research focuses on improving the detection of Medicare fraud using machine learning, particularly addressing the issue of class imbalance where fraudulent claims are significantly fewer than legitimate ones. A classification model is developed to differentiate between fraudulent and non-fraudulent Medicare claims, utilizing advanced algorithms such as Extreme Gradient Boosting (XGBoost), Adaptive Boosting (AdaBoost), LightGBM, Decision Trees, Logistic Regression, and Random Forest. To mitigate the class imbalance, the Synthetic Minority Over-sampling Technique combined with Edited Nearest Neighbors (SMOTE- ENN) is employed, SMOTE-ENN combines the Synthetic Minority Over-sampling Technique (SMOTE) with Edited Nearest Neighbors (ENN) to both over-sample the minority class and under-sample the majority class. Enhancing the dataset and improving model outcomes. Experimental results confirm that SMOTE-ENN effectively boosts the identification of fraud. Evaluations on both unbalanced and balanced data sets show improvements in key performance metrics, including accuracy, precision, recall, and F1-score. These findings suggest that integrating SMOTE-ENN with ensemble learning models offers a strong and effective strategy for detecting Medicare fraud.