- Code: source code of our implementation
- Data: some source files of datasets used in experiments
- Appendix.pdf:
- The motivation for introducing prominence modeling in graph structure learning;
- Complete proofs for all the theoretical results in the manuscript, including Propositions 1, and Propositions 2;
- Time complexity and space complexity analysis of our proposed node-scale spatial learning and feature-scale spatial learning modules;
- More supplementary experiments that demonstrate the effectiveness and rationality of our method.
-
Enter the "Code" folder
-
To get the DutchWind dataset:
python preprocessing_dutchwind.py
python preprocessing_dutchwind_getadj.py
- To get the BeijingMEO dataset:
python preprocessing_beijingmeo.py
python preprocessing_beijingmeo_getadj.py
- To get the LondonAQ dataset:
python preprocessing_londonaq.py
python preprocessing_londonaq_getadj.py
- To get the CN dataset:
python preprocessing_cn.py
python preprocessing_cn_getadj.py
- To get the Los dataset:
python preprocessing_los.py
- To get the LuohuTaxi dataset:
python preprocessing_luohutaxi.py
python A_main.py
python A_main_forecasting.py
- DutchWind: https://www.knmi.nl/nederland-nu/klimatologie/uurgegevens
- BeijingMEO: https://www.dropbox.com/s/jjta4addnyjndd8
- LondonAQ: https://www.dropbox.com/s/ht3yzx58orxw179
- CN: http://research.microsoft.com/apps/pubs/?id=246398
- Los: https://github.com/lehaifeng/T-GCN/tree/master/T-GCN/T-GCN-PyTorch/data
- LuohuTaxi: https://github.com/lehaifeng/T-GCN/tree/master/T-GCN/T-GCN-PyTorch/data