OPTIMIZED CLUSTER HEAD SELECTION IN GAUSSIAN GRID WSNS USING HYBRID ML FRAMEWORK

Abstract [English]

Wireless Sensor Networks (WSNs) deployed over Gaussian grids face critical challenges in energy efficiency and network longevity due to constrained node resources and spatial limitations. Traditional cluster head (CH) selection methods, such as LEGN, TEGN, and FOA-CH, individually address aspects of energy optimization, load balancing, and spatial coverage but often fall short in dynamic and heterogeneous network environments. This paper proposes a novel hybrid machine learning (ML) framework that combines features from these existing approaches to enable adaptive, energy-efficient, and fair CH selection. The framework constructs comprehensive feature vectors for each node, capturing residual energy, energy consumption ratio, historical CH frequency, spatial coordinates, and FOA-based optimization metrics. A predictive ML model dynamically selects candidate CHs, followed by localized Firefly Optimization Algorithm (FOA) refinement to improve spatial distribution and load balancing. Performance evaluation demonstrates improvements in network lifetime, average residual energy, load fairness, and data reliability, validating the effectiveness of the proposed approach in enhancing energy utilization and extending operational longevity of Gaussian grid-based WSNs.