Institutional Repository [SANDBOX]
Technical University of Crete
EN  |  EL

Search

Browse

My Space

Methodologies for the prediction of network usage within the context of cellular hotspots

Koutroumpas Georgios

Full record


URI: http://purl.tuc.gr/dl/dias/DA195DBC-1FFE-47AC-A174-6940063035C7
Year 2022
Type of Item Diploma Work
License
Details
Bibliographic Citation Georgios Koutroumpas, "Methodologies for the prediction of network usage within the context of cellular hotspots", Diploma Work, School of Electrical and Computer Engineering, Technical University of Crete, Chania, Greece, 2022 https://doi.org/10.26233/heallink.tuc.94673
Appears in Collections

Summary

Today we live in a society that relies upon technology. Everything we see around us has become more and more advanced with the addition of smart phones, smart cars and maybe even smart clothes. Many ofthose devices require the remote use of the world wide web to function properly. This fact,in conjunction with the ever increasing population in central living areas, creates severe issues to mobile service providers.Sudden demand bursts of their service can cause bottlenecks to the network infrastructure, resulting in performance issues to the cellular antennas. An interesting solution for this problem is forecasting when and where those performance drops will happen and re-calibrating the network parameters, effectively avoiding disaster. In this work I propose aneural network algorithm that will handle the forecasting task of those performance drops, referring to them as hotspots. To achieve this goal I am going to cooperate with the company Telefonica, which will provide essential information gathered from its networks antennas, as well as important feedback towards the final product. Using a combination of Gated Recurrent Units and Graph Convolution Networks the plan is to capture spatial and temporal dependencies that exist in the networks behaviour, effectively predicting most of the real performance drops in long prediction horizons. The focus of this work is to have accurate predictions of as many hotspots as possible and on the same time support a vast amount of antennas in the calculation.

Available Files

Services

Statistics