Table of Contents
Often the graph has to be extracted from data using some time-series similarity method (e.g., Pearson correlation, Granger causality, correntropy, etc). This project consists in benchmarking the effectiveness of the different similarity scores in extracting graphs that are useful for time series forecasting with GNNs.
- Python
- Anaconda
- Numpy
- Pandas
- Keras
- Tensorflow
- Spektral
- Matplotlib
- Seaborn
- scikit-learn
- scipy
- tqdm
- statsmodels
- fastdtw
In order to start the entire benchmark, is sufficient to execute the main.py file, which is set up to be ready to go.
If you want to try some other methods, is sufficient to add them on the main
Distributed under the MIT License. See LICENSE.txt for more information.
Gabriel Henrique Carraretto - carrag@usi.ch Michele Damian - damiam@usi.ch Riccardo Corrias - corrir@usi.ch
Project Link: https://github.com/gabecarra/GDL_Project