ALGORITMOS SUPERVISADOS PARA LA PREDICCIÓN DEL ANCHO DE BANDA DE LAS APLICACIONES EN AMAZON WEB SERVICE DESDE UNA PYME RURAL

Abstract

This article presents a methodology to measure the bandwidth behaviour by making predictions of the network traffic that connects to the cloud in small and medium enterprises in rural areas with difficult access in Colombia, in order to optimize network resources over time and ensure the quality of service in web applications. A comparative study of three neural network algorithms that model a multilayer neural network is performed, selecting the one that has a minimum error that approaches zero; the selected algorithm is trained from a data source to predict the network traffic that connects to the cloud.It is necessary to analyse network behaviour to ensure the quality of web applications in the cloud that transmit information such as data, images, sound, video, etc., some in real time, and that generate large volumes of traffic. Understanding the traffic flowing through the network enables network capacity planning when managing limited resources, such as in the case of small and medium-sized enterprises in rural areas. As a product of the research analysis, a free software prototype will be developed to perform the measurements and predictions in rural areas. The results of the implementation show that the proposed approach is superior to other forecasting methods in terms of accuracy and predictability.