University of Twente Student Theses
Examining the potential of Graph Neural Networks on road network data for traffic crash prediction
Doedens, W.J. (2025) Examining the potential of Graph Neural Networks on road network data for traffic crash prediction.
|
PDF
6MB |
| Abstract: | The study proposes training Graph Neural Networks on road network data to identify these factors for traffic crash prediction. Two machine learning (ML) models were trained to predict traffic accidents: a Multi- Layer Perceptron (MLP) and a Graph Convolution Network (GCN). Additionally, to address the class imbalance in traffic accident data, Class Balanced loss is introduced to give greater weight to minority classes. GNNExplainer, an Explainable Artificial Intelligence (XAI) method combined with partial dependence plots, are used in this study to determine which road factors are important for the ML models. The results show that it is harder for Graph Convolution Networks to classify minority classes correctly. Moreover, the results confirm that the chances of crashes increase as the number of cyclists and motorised vehicles on the road increases. The centrality of the road and its proximity to commercial facilities also increase the likelihood of crashes. Furthermore, increasing the amount of traffic lights decreases crash probability. Lastly, the findings show that separating bicycle lanes on 50 km/h roads and including bicycle lanes on 30 km/h roads increase traffic safety. |
| Item Type: | Essay (Master) |
| Faculty: | EEMCS: Electrical Engineering, Mathematics and Computer Science |
| Subject: | 31 mathematics, 55 traffic technology, transport technology |
| Programme: | Applied Mathematics MSc (60348) |
| Link to this item: | https://purl.utwente.nl/essays/105184 |
| Export this item as: | BibTeX EndNote HTML Citation Reference Manager |
Show download statistics for this publication
Show download statistics for this publicationRepository Staff Only: item control page