Mecanismo para el control de admisión de llamadas soportado en técnicas de aprendizaje automático

Abstract

El servicio de telefonía IP ha logrado gran auge en los últimos años al ser una solución económica especialmente para las empresas; sin embargo, aún existen algunos usuarios que prefieren la red telefónica conmutada por su excelente calidad. La aplicación de técnicas de inteligencia artificial al servicio de voz sobre IP promete disminuir esta brecha y propiciar la supervisión y mejoramiento de la calidad. Este estudio pretende hacer uso de técnicas de aprendizaje de máquina para analizar el estado de la red y proponer un control de admisión de llamadas mediante la premisa de políticas de calidad.

Among the convergent services, such as Voice over IP, it is necessary to address the challenges so that network administrators can guarantee the Quality-of-Service (QoS). Taking into account that the IP protocol works under the mechanism of the best effort, which does not guarantee quality of service in communications, one of the challenges is to model the operation of the service and capture the behavior of the network to propose a mechanism with QoS. This document proposes a mechanism based on quality of service policies, to improve the functions of network management through the application and comparison of Machine Learning (ML) techniques, taking advantage of predictive reasoning to make precise decisions for effective management. A case study is presented in a computer network for the application of these techniques to the problem of Call Admission Control (CAC). The study concludes that the use of ML techniques is a feasible solution when applied to the analysis of the QoS for the generation of CAC's policies.