TransNet

We introduce the Transport-informed Graph Neural Network (TransNet), a machine learning model that forecasts PM2.5 concentrations for +72-hours by learning coupled Advection-Diffusion-Reaction (ADR) operators, enabling efficient operational deployment.

Codes for TransNet: A Transport-Informed Graph Neural Network for Forecasting PM2.5 Concentrations across South Korea - Rijul Dimri, Yunsoo Choi, Deveshwar Singh, Jincheol Park, Nima Khorshidian

Scripts

The following scripts are provided:

• 📝: TransNet_train.py: TransNet model code.
• 📝: graph.py: Graph construction code.
• 📝: train_distributed.py: Distributed training code.
• 📝: evaluation.py: Evaluating code for year 2021.
• 📝: preprocessing.py: Preprocessing code that includes interpolation, Variational Mode Decomposition (VMD), and Principle Component Analysis (PCA).
• 📝: evaluation_notebook.py: Evaluating notebook.

Getting started :

To isntall the environment, we suggest using the environment.yml file when creating an environment.