Abstract
Abstract : The menace of fraud today cannot be underestimated. The healthcare system put in place to facilitate rendering medical services as well as improving access to medical services has not been an exception to fraudulent activities. Traditional healthcare claims fraud detection methods no longer suffice due to the increased complexity in the medical billing process. Machine learning has become a very important technique in the computing world today. The abundance of computing power has aided the adoption of machine learning by different problem domains including healthcare claims fraud detection. The study explores the application of different machine learning methods in the process of detecting possible fraudulent healthcare claims fraud. We propose a data mining model that incorporates several knowledge discovery processes in the pipeline. The model makes use of the data from the Medicare payment data from the Centre for Medicare and Medicaid Services as well as data from the List of Excluded Individual or Entities (LEIE) database. The data was then passed through the data pre-processing and transformation stages to get the data to a desirable state. Once the data is in the desired state, we apply several machine learning methods to derive knowledge as well as classify the data into fraudulent and non-fraudulent claims. The results derived from the comprehensive benchmark used on the implemented version of the model, have shown that machine learning methods can be used to detect possible fraudulent healthcare claims. The models based on the Gradient Boosted Tree Classifier and Artificial Neural Network performed best while the Naïve Bayes model couldn’t classify the data. By applying the correct pre-processing method as well as data transformation methods to the Medicare data, along with the appropriate machine learning methods, the healthcare fraud detection system yields nominal results for identification of possible fraudulent claims in the medical billing process.
M.Sc. (Computer Science)