Clustering of sensor nodes (SNs) is an unsurpassed energy management method in wireless sensor networks (WSNs) that ensures efficient energy balancing and duty-cycling, and improves the lifespan of the network by minimizing intra-cluster communication cost. Thus, since any incidences of misclustering shortens the lifespan of WSN, this paper presents an efficient, unbiased and more stable approach for evaluating the optimality of event-reporting (E-R) clusters in WSNs using the theory symbolic classifiers. Using realistic dataset derived from 1500 randomly deployed SNs, our results showed that the optimal number of clusters that guarantee optimal E-R accuracy and lengthened WSN lifespan by minimizing the intra-cluster communication costs are 240 clusters for classical K-Means method and 390 clusters for Extreme Learning Machine-Auto Encoder (ELM-AE). This method outperformed the classical inertia-based approach by establishing the optimal proxy E-R clusters which ensures higher E-R accuracy and energy efficiency of SNs. The experiment was done using realistic dataset extracted from randomly deployed 1500 SNs, and so our result is credible for the assessment of cluster qualities in other WSNs.
Effah, E., & Thiare, O. (2019). Estimation of optimal number of clusters: A new approach to minimizing intra-cluster communication cost in WSNS. International Journal of Innovative Technology and Exploring Engineering, 8(7), 521–524.