A Hybrid Approach for Network Lifetime Enhancement in Wsns with Graph Neural Networks and Probabilistic Regression

Abstract

Routing protocols energy consumption can heavily influence the network lifetime of a Wireless Sensor Network (WSN). Specifically, to reduce energy, data aggregation is utilized to discard data redundancy at each sensor and minimize the amount of data packet transmitted in a WSN. Moreover, energy-efficient routing is extensively utilized in deciding the optimal route between source and destination, to minimize energy for relaying the sensed data packets. Owing to the energy restrictions of the sensor nodes in WSN, employing an optimal model for routing and controlling WSNs can be efficient in improving energy efficiency and overall network lifetime. To overcome these issues, proposed Deep Correlated Graph Neural Network and Gaussian Probabilistic Regression (DCGNN-GPR) method introduced for network lifetime optimization in WSN. Initially, Deep Dominant Correlated and Rescorla Wagner Graph Neural Network are performed where determine the back propagation model lesser dominant and better correlated regions were considered. After that, Gaussian Probabilistic Regression models for network lifetime optimization is performed to obtain the minimum dominant highly correlated regions (i.e., highly correlated sensors) as input and improve the overall network lifetime by tradeoff sensor nodes of minimal dominant highly correlated regions. Finally, simulations were performed to evaluate the performance of the proposed network lifetime optimization method and compared it with that of the conventional methods for improving and optimizing network lifetime and discusses the trade-offs that exist between them. Lifespan of wireless sensor network based on the proposed method is greatly increased whereas the energy consumption, network life time and training time is greatly decreased by the techniques we have proposed.