Machine Learning Approaches for Privacy-Preserving Data Management in Cloud-Based Health Insurance

Abstract

Cloud-based health insurance systems require effective protection of sensitive user information while maintaining accurate and intelligent decision-making capabilities. However, existing approaches often face challenges related to privacy preservation, computational complexity, and reduced predictive performance under strict security constraints. This study proposes a privacy-preserving machine learning framework for secure data management in cloud-based health insurance environments. The framework integrates data cleaning, data masking, and secure feature transformation techniques to safeguard confidential information while preserving data utility. An optimized XGBoost classification model is employed to predict customer interest in insurance products. The proposed methodology includes missing value handling, duplicate removal, feature encoding, and secure dataset partitioning for model training and evaluation. Experimental results demonstrate strong predictive performance, achieving an accuracy of 97.7%, precision of 96.9%, recall of 97.4%, and F1-score of 97.1%, indicating a highly reliable and balanced classification system. The findings show that privacy-preserving transformations have minimal impact on predictive effectiveness while significantly enhancing data security. The proposed framework offers a practical, cost-effective, and scalable solution for secure cloud-based health insurance management, supporting privacy protection and efficient data-driven decision-making in modern healthcare insurance services.