A neural network-guided correlation thresholding technique for wireless sensor networks

Abstract

Wireless sensor networks comprise autonomous nodes monitoring environments while managing energy consumption. Extending network lifespan involves minimizing redundant transmissions by filtering identical information from correlated sensor nodes. Correlation depends on spatial locations and it quantifies sensor relationships. It allows nodes classification as correlated or uncorrelated using a threshold value between 0 and 1 which enhances network lifespan by 60–65 %. Selecting the optimal threshold considering network metrics remains a challenge. This study proposes a neural network-guided correlation thresholding technique (NNCT). NNCT aims to identify the optimal threshold for filtering redundant reports by training a neural network on varied parameters like node density, threshold, coverage, and events using different clustering methods. The process involves implementing spatial correlation models, virtual clustering, event detection, and iterative threshold-based node selection. The neural network achieves a high accuracy of R = 0.97813, offering a promising adaptive solution for automated threshold selection in correlation-based filtering.