LightGBM: Predicting Average Localization Error Through Particle Swarm Optimization in Wireless Sensor Networks

Localization of nodes represents a significant concern within Wireless Sensor Networks (WSNs). Predicting localization error accurately offers several advantages such as improved network performance, optimized resource allocation, and cost reduction. This paper presents a comprehensive investigation into the estimation of average localization error (ALE) in WSNs. The study employs the LightGBM algorithm, incorporating a particle swarm optimization (PSO) methodology to enhance predictive accuracy. The performance of the proposed model is assessed through comparison with various alternative machine learning methods, employing the R² and MSE evaluation metrics. The proposed LightGBM-PSO model outperforms the other methods with the R² value of 0.9043 and the MSE value of 0.1217. Moreover, feature importance analysis has been achieved. The number of iterations shows the greatest influence on the estimation of ALE, whereas the transmission range demonstrates the least impact.