A Prediction Model For Performance Analysis in Wireless Mesh Networks

Abstract

Analysis of computer networks is an important study field that must be handled carefully in order to make communication systems work properly. Efficient evaluation and remodelling of system according to factors affecting the performance is required. For this aim, many techniques have been proposed, so far. However, machine learning methods are getting more preferable than others with their cost-effective and faster solutions. In this study, generalized regression neural networks (GRNNs) approach was employed in order to predict the output, packets dropped of a sample DMesh network simulation. The simulation is driven by parameters such as number of nodes, number of gateways, number of channels used, and traffic density. It was observed that parameters: traffic density and number of channels used, have a direct impact on error rate of the regression model. The high variance explained values show that GRNN approach can represent real characteristics of DMesh architecture.