DATA DRIFT DETECTION AND MODEL RELIABILITY MANAGEMENT IN PRODUCTION HEALTHCARE MLOPS

Abstract

Healthcare organizations increasingly rely on machine learning models for clinical analytics, claims prediction, resource optimization, and operational decision-making. However, maintaining model reliability in production environments remains challenging due to changing data distributions, evolving healthcare practices, regulatory updates, and dynamic patient or operational patterns. This paper proposes a Data Drift Detection and Model Reliability Management framework for production healthcare MLOps by integrating machine learning operations, automated monitoring, statistical analysis, explainable artificial intelligence, cloud computing, and intelligent data pipelines. The proposed methodology enables continuous detection of data drift, concept drift, model degradation, performance variations, and operational risks while supporting automated model validation and retraining workflows. Experimental evaluation demonstrates improvements in model stability, prediction reliability, monitoring efficiency, compliance management, and healthcare analytics performance. The proposed framework provides a scalable and intelligent solution for maintaining trustworthy AI systems in healthcare environments through continuous monitoring, adaptive learning, and automated reliability management. Keywords— Data Drift Detection, Model Reliability, Healthcare MLOps, Machine Learning Operations, Predictive Analytics, Explainable AI, Model Monitoring, Healthcare AI.