A multi-phase task scheduling with NSGA-II integrated ant colony optimization in fog-cloud environment

Efficient task scheduling in fog-cloud environments has become a critical challenge in modern healthcare systems due to the increasing demand for data-intensive medical applications and real-time service delivery. Managing computational tasks under heterogeneous resource constraints while maintaining balanced workload distribution, low latency, reliability, and high system performance remains a significant issue in healthcare fog-cloud infrastructures. To address these challenges, this paper proposes a prediction-aware multi-layer intelligent scheduling framework for healthcare fog-cloud environments. Initially, a rule-based scheduling layer performs preliminary task filtering based on deadline, budget, VM resource availability, and healthcare-specific constraints to generate feasible task-to-VM mappings. Subsequently, a Long Short-Term Memory (LSTM)-based prediction model forecasts near-future virtual machine workloads using historical workload observations and temporal dependencies, enabling prediction-aware and adaptive resource allocation. Finally, a hybrid optimization framework combining Ant Colony Optimization (ACO) and Non-Dominated Sorting Genetic Algorithm II (NSGA-II) generates near-optimal scheduling solutions by jointly minimizing execution cost, makespan, and latency while maximizing throughput and resource utilization. Under multi-objective constraints, experimental evaluation using synthetically generated healthcare-oriented workloads demonstrates that the proposed framework achieves an execution cost of 25,842.61 and a makespan of 21,846.73. The framework further achieves 98.40% emergency task priority satisfaction, 98.90% reliability, and 98.10% fault tolerance, while reducing the critical deadline violation penalty to 2.15. Compared with the corresponding framework without LSTM-based workload prediction, the proposed approach reduces execution cost by 12.37%, makespan by 4.91%, and average latency by 73.93%, while improving throughput by 89.15% and the resource utilization score by 82.69%. Comparative analysis against ACO, NSGA-II, Grey Wolf Optimization (GWO), Hummingbird Algorithm (HBA), and NSGA-III demonstrates the effectiveness of the proposed framework across multiple scheduling and healthcare-oriented performance measures. Statistical analysis using the Wilcoxon Signed-Rank test, paired t -test, Friedman test, and ANOVA further supports the statistical robustness and effectiveness of the proposed framework.