Analysis of Wireless Traffic Data through Machine Learning

The paper presents an analytical study on a wireless traffic dataset carried out under the different approaches of machine learning including the backpropagation feedforward neural network, the time-series NARX network, the self-organizing map and the principal component analyses. These approaches are well-known for their usefulness in the modeling and in transforming a high dimensional data into a more convenient form to make the understanding and the analysis of the trends, the patterns within the data easy. We witness to an exponential rise in the volume of the wireless traffic data in the recent decade and it is increasingly becoming a problem for the service providers to ensure the QoS for the end-users given the limited resources as the demand for a larger bandwidth almost always exist. The inception of the next generation wireless networks (3G/4G) somehow provide such services to meet the amplified capacity, higher data rates, seamless mobile connectivity as well as the dynamic ability of reconfiguration and the self-organization. Nevertheless, having an intelligent base-station able to perceive the demand well before the actual need may assist in the management of the traffic data. The outcome of the analysis conducted in this paper may be considered in designing an efficient and an intelligent base-station for better resource management for wireless network traffic