Decisions in Structure
Ya Song developed GNN methods that model temporal and spatial patterns in industrial data, with applications in maintenance and combinatorial optimization.
Ya Song successfully defended his PhD on September 24, 2025, at the Department of Industrial Engineering & Innovation Sciences at 黑料福利网. His research, conducted within the Information Systems (IS) group, investigates how Graph Neural Networks (GNNs) can support better industrial decision-making by modeling temporal and spatial dependencies in data.
Industrial Context
In sectors like manufacturing and logistics, decisions often rely on complex patterns across time and space. Examples include predicting maintenance needs from sensor data or selecting the right algorithm for routing tasks. Song demonstrates that conventional models struggle to capture these relationships, which can lead to inaccurate predictions.
Asset Prognostics
Entrepreneurs and organizations face variable data, irregular monitoring sequences, and structural complexity. In maintenance processes, for instance, it鈥檚 difficult to pinpoint the right moment to intervene. Song developed methods that better capture degradation patterns and sensor relationships, resulting in more robust predictions that align with real-world conditions.
Algorithm Selection
In combinatorial optimization tasks like the Traveling Salesman Problem, the research also offers valuable insights. By analyzing local spatial patterns and ensuring geometric consistency, the model helps identify the most suitable algorithm for each scenario. This supports more efficient planning and better adaptation to changing operational demands.
Ya Song defended his thesis on September 24, 2025.
Thesis title: .
Supervisors: Yingqian Zhang and Laurens Bliek.