Establishing a machine learning model for fraud detection in medical insurance claims

Abstract

Insurance fraud is a growing issue that has an impact on members and businesses. Numerous efforts have been made throughout the years to identify fraudulent claims early on and before paying out the claims. However, it has not been demonstrated that the techniques are enough to address the issue. The capabilities of machine learning techniques have greatly increased, and they can now handle challenging issues. The identification of false claims has been explored and used using a variety of Machine Learning techniques. However, not all Machine Learning techniques have been looked at. In the field, certain algorithms are still a mystery. In the project that follows, the K-Nearest Neighbor model, decision trees, logistic regression, and Nave Bayes Classifier will be created, examined, and compared. The Nave Bayes algorithm is the one that is used for fraud detection the most often out of the three. After developing the models using the data that was provided, a comparison of the models revealed that Nave Bayes Classifier is the most effective algorithm for fraud detection in the insurance business.