Machine Learning-Based Fraud Detection in Healthcare Insurance with Optimized Storage in Blockchain Using Zero-Knowledge Proofs

Abstract

The prevalence of fraudulent activities in the insurance industry is alarmingly increasing and requires innovative solutions. The primary objective of the research is to identify instances of fraudulent insurance claims through machine learning algorithm and develop an efficient storage system of the insurance claims, which is secured, private and suitable for the industry. The XGBoost model with SMOTE oversampling is proven to be distinguished among other models. Hyperledger Fabric, a permissioned ledger is used to store and retrieve the insurance data, ensuring the reliability, immutability and authorization. To protect the data from public visibility, zero knowledge proof technology is utilized. This ensures the privacy and authenticity of the information. Overall, this research provides an efficient and automated claim validation system with higher accuracy and efficient storage in blockchain while preventing information leak.