| KEYWORD |
Theory enhanced traffic prediction using Graph Neural Networks
keywords DATA MODELLING, GRAPH NEURAL NETWORKS, MACHINE LEARNING, PREDICTION METHODS
Reference persons DANIELE APILETTI
External reference persons Simone Monaco
Research Groups DAUIN - GR-04 - DATABASE AND DATA MINING GROUP - DBDM
Thesis type ANALISI DATI, ANALITICA E SPERIMENTALE, MODELLAZIONE E ANALISI DATI
Description Traffic congestion is a significant problem in urban areas, causing delays and increased air pollution. Accurate traffic prediction can help alleviate these issues by providing information for better traffic management and route planning. This thesis proposes a novel approach to traffic prediction using graph neural networks and domain theory injection.
Graph neural networks (GNNs) have shown great promise in modeling complex systems with graph structures, such as road networks. On the other hand, domain theory injection in deep learning models is a promising technique that allows the design of more stable and explainable algorithms.
Methodology: In this thesis, we will develop a traffic prediction algorithm that combines GNNs with domain injection. We will use a GNN to model the road network and traffic flow and inject domain knowledge about traffic patterns and equations behind the flow into the model.
Expected Outcomes: The algorithm will provide more accurate predictions than existing methods. Moreover, we are interested in investigating the potentialities of theory-guided deep learning models and their generalization to broader scenarios.
Prerequisites: Strong computer science and mathematics background and good Python programming skills. Knowledge of machine learning, deep learning, and graph theory would be beneficial. Additionally, experience with relevant tools and libraries for implementing machine learning algorithms (e.g., PyTorch) would be highly appreciated.
Deadline 14/06/2024
PROPONI LA TUA CANDIDATURA