The rapid advancement of machine learning (ML) technologies has opened new frontiers in healthcare analytics, particularly in the domains of real-time health risk stratification and cost prediction modeling. Traditional actuarial and statistical methods, while foundational, often lack the agility and precision necessary to manage increasingly complex patient populations and dynamic health system demands. The integration of advanced ML-driven business analytics offers a transformative pathway, enabling healthcare organizations to anticipate clinical risks, allocate resources more efficiently, and transition toward proactive, value-based care delivery models. This paper explores the deployment of advanced ML algorithms—such as gradient boosting, deep learning, and ensemble techniques—for predictive health risk stratification at the individual and population levels. It examines how real-time analytics platforms, powered by multimodal patient data and financial indicators, enable the continuous refinement of cost prediction models, leading to more accurate budgeting, targeted interventions, and risk-sharing contract strategies. Emphasis is placed on the critical role of feature engineering, data governance, model interpretability, and ethical considerations, particularly around algorithmic bias and transparency. Drawing on case studies from leading integrated health systems and payer organizations, the paper demonstrates how predictive insights derived from ML models can reduce hospital admissions, optimize care management programs, and improve financial forecasting accuracy. By narrowing from the broader context of ML's impact on healthcare to specific applications in risk and cost prediction, this study provides actionable frameworks for integrating advanced analytics into strategic healthcare operations and policymaking in an era of rising complexity and accountability.