A Distributed Data Fusion Approach to Mitigate Node Redundancies in an IoT Network

Data redundancy is one of the prominent causes of excess energy expenditure and network lifetime degradation in the Internet of Things (IoT)-based sensor networks. Generally, the wireless sensor network (WSN) is equipped with low computational capability nodes which are vulnerable to the data redundancy issue when dealing with the raw data. We introduce a model in which the data sensed by the sensors containing redundancies are analyzed using Kernel density estimation (KDE) in combination with Pearson’s divergence with the distributed approach to calculate the global probability density function between the data points of two consecutive periods. The final data without redundancies are fused to the sink node using bacterial foraging optimized extreme learning machine (BFO-ELM) which also classifies the redundant nodes. The simulation performed on three real-world datasets shows that the proposed technique records the accuracy of about 95% and the energy consumption reduction by approximately 30% and promises a high network lifetime when compared with other distributed and centralized methods.