Implementation of XGBoost Method for Healthcare Fraud Detection

Abstract

The health care systems are quickly adapting digital health records, which will exponentially increase the quantity of medical data. The systems are generally faced with unsustainable costs and large volume of electronic medical data. Therefore, more efficient research, practices and real-world applications are needed to take advantage of all benefits of medical data. One strategy to cut back on the rising costs is the detection of fraud. In this paper, Xgboost, which is an implementation of gradient boosted decision trees, is employed, along with more traditional supervised algorithms to include Random Forest, k-Nearest neighbor. The List of Excluded Individuals/Entities (LEIE) database, which contains excluded providers information, is used to label as a fraud in the Medicare Part B dataset. Thus, the data has become available for use with supervised methods. According to the experimental results, the XGBoost algorithm outperformed traditional machine learning algorithms in terms of performance.