Network monitoring and performance assessment: from statistical models to neural networks
EntityUAM. Departamento de Tecnología Electrónica y de las Comunicaciones
SubjectsComputer networks; Data modeling; Functional data analysis; Telecomunicaciones
NoteMáster en Investigación e Innovación en Tecnologías de la Información y las Comunicaciones
Esta obra está bajo una licencia de Creative Commons Reconocimiento-NoComercial-SinObraDerivada 4.0 Internacional.
In the last few years, computer networks have been playing a key role in many different fields. Companies have also evolved around the internet, getting advantage of the huge capacity of diffusion. Nevertheless, this also means that computer networks and IT systems have become a critical element for the business. In case of interruption or malfunction of the systems, this could result in devastating economic impact. In this light, it is necessary to provide models to properly evaluate and characterize the computer networks. Focusing on modeling, one has many different alternatives: from classical options based on statistic to recent alternatives based on machine learning and deep learning. In this work, we want to study the different models available for each context, paying attention to the advantage and disadvantages to provide the best solution for each case. To cover the majority of the spectrum, three cases have been studied: time-unaware phenomena, where we look at the bias-variance trade-off, time-dependent phenomena, where we pay attention the trends of the time series, and text processing to process attributes obtained by DPI. For each case, several alternatives have been studied and solutions have been tested both with synthetic data and real-world data, showing the successfulness of the proposal
Google Scholar:Perdices Burrero, Daniel
This item appears in the following Collection(s)
Showing items related by title, author, creator and subject.