Metrologia na Era do Aprendizado de Máquina

Abstract

The growth in the number of mobile and cellular networks arouses interest in end-to-end performance measurements of these networks and their impact on mobile applications [15]. Among these measurements, we find those based on the quality of the user experience, which is an excellent source of information about network management. In this context, the implementation of Artificial Intelligence and Machine Learning increases the efficiency of the monitoring process [21]. In this short course, we will discuss how we can use objective parameters of quality of services such as delay, throughput, packet loss, and jitter to estimate and predict the quality of user experience using Machine Learning [21]. Furthermore, it is possible to use Artificial Intelligence to estimate the multiple indicators of the quality of user experience, even in a network that adopts end-to-end encryption [49,27]. In addition, Machine Learning methods help acquire knowledge about the functioning of networks, understand how the structures of existing networks are, and allow the establishment of new networks with less effort. Among the several Network Metrology problems that can benefit from Machine Learning, we can mention network attack detection, user experience quality prediction, application anomaly detection, and network failures.