Detecting and Combating Fraudulent Health Insurance Claims Using ANN

Abstract

This work was funded by the National Nature Science Foundation of China (71774069), 2014 “Six Talent Peaks” Project of Jiangsu Province (2014- JY-004)

Abstract

While governments and private sector stakeholders are taking steps to improve the access and quality of health care service to its citizenry, a lot of resources are lost every year due to fraudulent health insurance claims. The aim of this paper is to explore a more robust and accurate ways of predicting fraudulent health insurance claims by the use of artificial neural network (ANN). Using the fraud diamond theory (FDT)’s fraud elements as fraud indicators, a fraud prediction model was created to determine whether a claim presented by a subscriber (individual) is fraudulent or non-fraudulent by varying severally the number of epoch, hidden layer number and threshold of the artificial neural network on a 14 input data to obtain an optimal parameter for the model.The model was able to predict accurately 98.98% with an MSE of 0.0086, which outperformed other artificial neural network (ANN) methods used to predict fraudulent health care claims. The incorporation of the capacity indicator of the fraud diamond theory (FDT) makes this model a tool not only for prediction but also pre-empting the occurrence of fraud. This study is the first to adopt the fraud diamond theory’s fraud elements as fraud indicators together with artificial neural network (ANN) in predicting fraudulent health insurance claims.

Keywords: health insurance claim, ANN, fraud prediction model, fraud diamond theory