Congestion detection in wireless sensor networks using MLP and classification by regression

Abstract

Wireless Sensor Network (WSN) is network of hundreds or thousands of sensors. Congestion occurs in wireless sensor networks when all the sensors nearby event start sending data to the base station. Congestion results in less throughput and non reliability of a system. The machine learning algorithms can be applied for congestion detection in network and then congestion can be mitigated by lowering the transmission rate. In this paper we analyze the performance of multilayer level perception (MLP) — a neural network technique and classification by regression algorithms. The machine learning techniques are applied to detect the different levels of congestion in as low, medium or high. It is found that classification by regression is more efficient than MLP in detecting the congestion for the generated data set of WS'N simulation using NS2.