CASCADED PROPENSITY MATCHED FRAUD MINER: DETECTING ANOMALIES IN MEDICARE BIG DATA

CASCADED PROPENSITY MATCHED FRAUD MINER: DETECTING ANOMALIES IN MEDICARE BIG DATA

Healthcare fraud is a serious cause of concern due to its unrestrained growth in funded medical aid plans around the globe.Apart from the monetary deficiencies caused by fraudulent practices, a greater challenge is the shortage of leftover funding thattranslates into unavailability of medical services for the ones who need it the most. Organizations such as the Center for MedicareServices (CMS) in the U.S. have started providing access to comprehensive medical big data to face the onslaught of healthcarefrauds. Through the use of statistical machine learning and the ability to process medical big data, we are starting to see promisingdevelopments for the analysis of fraud in these expansive medical databases. This paper builds upon our previous work in fraudtype classifications and the multidimensional Medicare data to provide a multivariate data model that aids in predicting thelikelihood of healthcare fraud instances. A novel Cascaded Propensity Matching (CPM) Fraud Miner is proposed to identifyfraudulent outliers in the CMS Medicare dataset. The proposed CPM Fraud Miner targets the most widespread known types ofmalpractices and should be helpful in exploring new and evolved fraud practices. This paper also performs a comprehensivereview of current state-of-the-art models in healthcare fraud and functionality evaluation against leading methods with knownfraud cases.