Missing data recovery using reconstruction in ocean wireless sensor networks

Abstract

Ocean Wireless Sensor networks (OWSNs) usually operate under adverse physical conditions and are not very reliable. In addition, the threat for security adds to adversity. Large-scale data loss occurs frequently owing to data packet collision, signal attenuation, wave shadowing and malicious attacks in OWSNs. In this paper, we propose a novel data recovery using reconstruction algorithm based on improved K-means algorithm and PSO-RBF (radial basis function neural network optimized by particle swarm optimization algorithm) Neural Network (KPR-NN) to predict the missing data for sensors in OWSNs. In this approach, we use a node clustering module and data recovery module to reconstruct the missing data. We first generate clusters according to the improved K-means in the node clustering module, and then in the data recovery module, PSO-RBF Neural Network is applied to reconstruct the missing data. Simulation results demonstrate that our proposed approach for missing data recovery in OWSN outperforms the selected benchmark algorithms in terms of accuracy. Simultaneously, it can effectively reduce communication cost and prolong the lifetime of OWSNs by using the predicted values as a replacement for the real values in cluster head nodes. OWSN shows strong spatial–temporal relations in the data collected and hence better accuracy is achieved by employing the proposed algorithms.