Μεταπτυχιακές Εργασίες

Μόνιμο URI για αυτήν τη συλλογήhttps://pyxida.aueb.gr/handle/123456789/7

Περιήγηση

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

Α Β Γ Δ Ε Ζ Η Θ Ι Κ Λ Μ Ν Ξ Ο Π Ρ Σ Τ Υ Φ Χ Ψ Ω

Τώρα δείχνει 1 - 1 από 1

Assessment of a hidden Markov model for detecting abrupt shifts in time series of telecommunication data
(12/14/2021) Maragkopoulos, Giorgos; Μαραγκόπουλος, Γεώργιος; Athens University of Economics and Business, Department of Informatics; Ntzoufras, Ioannis; Vassalos, Vasilios; Vrontos, Ioannis
In recent years, internet and telecommunications made it possible for people around the world to connect, work and study from anywhere, at any time. However, to keep providing these benefits in a quality level, companies need to control the volume of signal power necessary in each day. This thesis proposes two models, first an abrupt shift capture model to detect aggressive movements based on a balanced – bagging multinomial logistic regression model, trained on labels produced by a hidden Markov model. In the second half, a time series regression model which forecasts the value of the next day using the n-beats model. The advantage of these models is quickly capturing relationships even in small datasets from patterns existing between time-lags. The experimental results show that the first model outperforms other baselines that were tested in terms of adjusted F1 scores and the same is true for the second model, in terms of regression scores and accuracy. Thus, they can potentially serve as classifiers / regressors in an automated tool which informs the company.