A Framework for Link Prediction in Urban Road Networks Based on Graph Theory and Machine Learning
Bildiriler > Tam Metin Bildiri
Yazarlar: Ercan GÜNBİLEK, Züleyha AKUSTA DAĞDEVİREN, Hasan BULUT
2025 9th International Artificial Intelligence and Data Processing Symposium (IDAP), , , Uluslararası, Tam Metin Bildiri, ss 1-8
DOI: 10.1109/idap68205.2025.11222319
Özet
The analysis of road networks is essential for a number of reasons, particularly in the context of urban planning, transportation management, cost reduction, disaster management and the general functioning of cities. In this study, we provide a comprehensive analysis that examines the network structures between European cities and evaluates the relationships between these cities using graph theory and machine learning techniques. The basis of the study is a network graph that represents the connections between European cities. This graph models the relationships between specific cities as edges and nodes and analyzes the importance of these nodes and edges in the network using various measures of centrality. These analyses include degree centrality, closeness centrality and betweenness centrality. A machine learning model is developed that takes into account the similar characteristics of cities by applying community detection methods to the graph. In this machine learning approach, different classification algorithms, including Random Forest, Gradient Boosting and Multi-Layer Perceptron (MLP), were comparatively evaluated for the link prediction task. The model with the best performance, MLP, was then used to predict potential new links based on the current network structure. Another important aspect of the study is the visualization of the results obtained. Using an interactive map showing the connections between the cities, the nodes and edges are color coded according to the centrality measures. These visualizations allow a better understanding of the structure of the road network and its important features.