Institutional Repository of Vilnius University
Modelių ir duomenimis grindžiamų prieigų palyginimas laiko-erdvės kelių tinklo modeliavime pritaikant grafų neuroninius tinklus /
Abstract
dc:descriptionModeling of traffic flow has become an integral part of urban life and has received much attention in the technology sector. Most architectural solutions for road network modeling are derived from deep learning techniques. Graphical neural networks are very effective in predicting traffic flows as they can model spatial relationships and adapt to data that have temporal and spatial characteristics. In this work, we explore the potential of applying graph neural networks to this problem. We develop a road network simulation model based on the architecture of the A3T-GCN model, i.e. a graph convolutional network with temporal attention, and real-world data, i.e. sensor readings from the Vilnius city road network. The prediction results of this model were compared with statistical methods and other deep learning methods. The comparisons confirmed the advantages of this model architecture in modeling road networks. The ability of the model to interpolate traffic flow data was not validated due to a possible inadequate interpolation implementation. The work concludes that graph neural networks are effective in modeling road networks in space and time.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2025
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Brazauskas, Deividas,
Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- lit
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD239029945&prefLang=en_US