Hello, I'm quite new to the forecasting domain and have a few questions regarding your work.
First, regarding reproduction: after running your script with the default settings, the data generated by the IGSTGNN model under a 6-hour prediction window shows a large variance from the expected results. I suspect there might be an issue with the step setting. Could you help clarify this?
I also have a few general questions:
Model selection: Why did you choose baseline models primarily designed for traffic prediction when dealing with a meteorological dataset? Did you fine-tune these models specifically for weather data?
Model complexity: You noted that simpler models performed better regarding error rates. Could you explain the reasoning behind this?
Code availability: I couldn't find the public code for IGSTGNN. Is it a custom implementation that you wrote yourself?
Thanks in advance for your help and for the great work!
Hello, I'm quite new to the forecasting domain and have a few questions regarding your work.
First, regarding reproduction: after running your script with the default settings, the data generated by the IGSTGNN model under a 6-hour prediction window shows a large variance from the expected results. I suspect there might be an issue with the step setting. Could you help clarify this?
I also have a few general questions:
Model selection: Why did you choose baseline models primarily designed for traffic prediction when dealing with a meteorological dataset? Did you fine-tune these models specifically for weather data?
Model complexity: You noted that simpler models performed better regarding error rates. Could you explain the reasoning behind this?
Code availability: I couldn't find the public code for IGSTGNN. Is it a custom implementation that you wrote yourself?
Thanks in advance for your help and for the great work!