Analytical Study on Fraud Detection in Healthcare Insurance Claim Data Using Machine Learning Classifiers

Abstract

Prediction of medical insurance fraud has become an active research topic. It is a demanding task among the business and academic researchers. The anonymous activities prevailing in the insurance claims have affected the financial growth of the health insurance companies. In this paper, an analytical study is done on the conventional Machine Learning (ML) classifiers like k-mean clustering, Support Vector Machine (SVM) and Naive Bayes (NB) are analyzed on a healthcare provider fraud detection dataset. The collected data is pre-processed by filtering techniques. The detection of fraudulent claims is achieved from the diagnosis and the total charged amount on a claim given by the different providers. Thus, the proposed claim data consist of claim amount; diagnosis and the providers. By using diagnosis attribute, as control (or) decision variables and the provider attribute as target class, the performances of the considered ML classifiers are presented in terms of classification accuracy, recall and precision. In addition to that, False Positive Rate (FPR) is also analyzed to evaluate the classifiers.