Efficiency of Naïve Bayes Technique Based on Clustering to Detect Congestion in Wireless Sensor Network

Abstract

Wireless sensor network (WSN) is the network of sensor nodes set up to supervise physical observable fact. Congestion is state in the network when too many packets are present in the network than capacity of network. Congestion can be at node level or link level. Our work is to related to node level congestion. Because of funnel like topology of wireless sensor network, the congestion occurs at the nodes near sink as all the nodes start sending data to sink node whenever an event occurs. Congestion detection is vital as it leads in poor performance of network. In this paper we have implemented the machine learning techniques to detect congestion in wireless sensor network using Ensemble approach of clustering and classification. The Naïve Bayes classification based on the K- means and Expectation- Maximization clustering algorithms are applied to generate the classifier model. The classification model is also generated using only Naïve Bayes algorithm and the performance is compared with classifier of ensemble approach. The analysis of performance parameters of the generated models indicates that EM based Naïve Bayes classifier model is more accurate in detection of the congestion for the generated our network data set.