A hybrid machine learning and metaheuristic optimization framework for energy-efficient data aggregation in real-time for rural IoT networks

Rapid deployment of Internet-of-Things (IoT) devices has led to a surge in real-time data generation, intensifying challenges related to energy consumption, bandwidth limitations, and network congestion. Traditional transmission methods suffer from decreased packet delivery ratios and long end-to-end delays. This work proposes a novel data aggregation model that addresses spatial and temporal redundancy while optimizing network parameters. Initially, a clustering approach employs K-Means to analyze spatial data patterns, refined using an Exponential Weighted Moving Average (EWMA). Cluster head (CH) selection is energy-aware to extend network longevity. A synergistic metaheuristic method, integrating Grey Wolf Optimization (GWO) with Greedy Perimeter Stateless Routing (GPSR), determines the optimal routing path from the CH to the sink node. Designed for rural and agricultural IoT networks, the proposed method achieves an average energy efficiency improvement of 11.1% over DA-MOMLOA, 4.7% over MOCRAW, and 20.9% over MOEA. After 2000 simulation iterations, the proposed model retains 40% of nodes alive, indicating significantly enhanced network longevity. It also improves the packet delivery ratio (PDR) by 2.1% over DA-MOMLOA, 1.3% over MOCRAW, 4.6% over MOEA, and 4.3% over LEACH, achieving a 97.32% PDR at high node density. Simulations in NS-3 confirm the model’s superior efficiency, reliability, and scalability in real-time IoT deployments.