A New Hybrid RNN-LSTM Model for Predicting the Health Insurance Prices

Health care expenditures are currently rising and will keep doing so in the future because of the growing incidence of lifestyle diseases. As the health insurance policy offers large financial coverage in the crucial medical emergencies, people always prefer taking it. But before signing up to a policy, it is important to predict the amount of medical expense which can be incurred in the future based on the lifestyle that a person leads. By accurately predicting the health insurance amount, the person can sign up to the best suitable policy and be benefitted from it. Also, this prediction can be alarming to few people who can later decide to change their lifestyle. To make accurate predictions, we have done a comparative analysis of deep learning and machine learning models and evaluated with various evaluation metrics such as Mean Squared Error, R2 Score and Cross Validation score. Usage of Flask, JavaScript, Docker and Azure along with the chosen Deep Neural Network (DNN) model ensures the end-to-end solution for the study. The end users of this system can be of two types, either the user is having their own past medical record, or the user does not have the track of their medical expense and wants to predict the amount based on the charges incurred for other people leading similar lifestyle. The first type of users can be handled with the hybrid model of Recurrent Neural Network (RNN) and Long Short-Term Memory (LSTM) model to identify the pattern in the given record to forecast their future monthly medical expense.