A HYBRID CNN-BIGRU MODEL WITH ATTENTION FOR PROACTIVE NETWORK PERFORMANCE ANALYSIS IN UNDERWATER WIRELESS SENSOR NETWORKS
DOI: https://doi.org/10.65725/JCISE/1/2/002
JOURNAL OF COMPUTATIONAL INTELLIGENCE SCIENCE AND ENGINEERING (JCISE)
ISSN: 3107-8168
Volume 1 Issue 2, October 2025

Abstract
Underwater Wireless Sensor Networks (UWSNs) are pivotal for oceanographic exploration, disaster prevention, and environmental monitoring. However, their performance is severely challenged by the unique characteristics of the aquatic medium, including high propagation delay, limited bandwidth, and dynamic node mobility. Traditional analytical models often fail to capture the complex, non-linear relationships governing network behavior. This paper proposes a novel deep learning framework for proactive network performance analysis in UWSNs. The model integrates Convolutional Neural Networks (CNN) for spatial feature extraction from node topology and Bidirectional Gated Recurrent Units (BiGRU) augmented with an attention mechanism to learn long-term temporal dependencies in network traffic and channel conditions. We evaluate our model on a comprehensive dataset simulating various underwater scenarios. The proposed Hybrid CNN-BiGRU-Attention model is benchmarked against several state-of-the-art algorithms. Results demonstrate its superior performance in accurately predicting key performance indicators (KPIs) such as end-to-end delay, packet delivery ratio, and network throughput. Our model achieves an average accuracy of 98.2% and an F1-score of 97.8%, significantly outperforming existing methods, thereby providing a robust tool for network optimization and fault prediction in challenging underwater environments.

Authors: A.Gobinath, A.Afrin Safna, and A.Anbarasi

Keywords: Underwater Wireless Sensor Networks; Deep Learning; Convolutional Neural Network; Bidirectional Gated Recurrent Unit; Attention Mechanism; Network Performance Prediction; Proactive Management.