GCRDD
This is the code for paper Graph Convolution Recurrent Denoising Diffusion Model for Multivariate Probabilistic Temporal Forecasting
Project Overview
The framework generally extends the probabilistic diffusion method to spatial-temporal forecasting with predefined topology information. In summary, our main contributions are as follows:
– Limited studies consider the structure information in the probabilistic multi-variate time series forecasting model . To the best of our knowledge, this is one of the pioneer applications of the temporal diffusion model with graph knowledge.
– We propose a modified graph diffusion module that was refined from the TimeGrad to capture hidden spatial dependencies among temporal features. Our method exploits a new field for diffusion models to handle spatial-temporal data.
– Comparing our method with state-of-the-art methods on two spatial-temporal traffic datasets demonstrates its competitive performance
Installation Instructions
See setup.py
Usage
python run.py config.ini
License
Specify the license information for your project. This helps others understand the permissions and restrictions for using your project.