Hybrid Machine Learning Model for Automated Health Insurance Claim Processing

Abstract

The health insurance sector processes millions of claims daily, requiring accuracy, fairness, and efficiency in claim approval and payout estimation. Traditional manual and rule-based claim evaluation systems are prone to bias, redundancy, and delays, resulting in financial and reputational losses. This paper presents a Hybrid Machine Learning (ML) Model integrating Random Forest (RF) and Linear Regression (LR) within a unified data preprocessing and prediction pipeline. The RF classifier determines claim approval likelihood, while LR estimates the expected claim payout. Data preprocessing includes missing value imputation, categorical encoding, normalization, and class rebalancing through Synthetic Minority Over-Sampling Technique (SMOTE). Experimental results on real-world datasets demonstrate an accuracy of 98%, precision of 0.97, recall of 0.98, and an F 1 score of 0.975, significantly outperforming standalone models. The proposed model offers an interpretable, scalable, and explainable framework for automated health insurance claim decision-making.

Ähnliche Arbeiten