Fraud Detection in Health Insurance Claims using Machine Learning and Deep Learning Techniques

A claim is what a doctor or hospital submits to the insurance company so they can get paid. A health insurance claim is a request that a health insurance policyholder submits to the Insurance Company in order to get the services that are covered in their health insurance policy.To tackle the problem of fraud in the medical insurance claims, health insurance companies make use of traditional rule-base models, but these models do not suffice anymore due to several factors such as the large volume of claims to be processed which makes the medical billing process prone to error, slow and sometimes inefficient. Hence there is a need to develop a novel fraud detection model for insurance claims processing. In this paper we are using the power of Machine Learning(ML) and Deep Learning(DL) algorithms to efficiently identify the fraudulent health insurance claims from the healthcare transaction dataset. Here we have considered a data set consisting of both legitimate and fraudulent claims. Initially we have implemented ML algorithms like SVM(Support Vector Machine), Logistic regression and Random forest for detection of anomalies and classification of health insurance claims into legitimate and fraudulent claims. Also DL algorithm MLP (Multilayer perceptron) is implemented. We have measured various metrics like Accuracy, Recall, Precision etc., for evaluating the performance of these ML and DL algorithms. All the models are good, but comparatively MLP has better performance with highest accuracy of 88.6 %. Study is also performed by varying the number of fraudulent records for the initial data set. This model helps in easy detection of the occurrence of fraudulent activities also is capable of fair fraud identification when applied to datasets with different fraud occurrence rates. The model can be used for decreased fraudulent activities in health insurance claims, it can become an efficient and effective application for Insurance clients and companies.