Looking at the date range in train and test shows that the submissions are not evaluated in future time but over the same period. Note that we don't know how the private data is split but I assume it's similar to the test data used for the public score.
Date time features are probably useful but that's different from modelling the problem as a time-series forecasting.
Be careful, there is a huge difference between time series forecasting and time series regression according to sktime website. I think this problem is more about time series regression.
I haven't considered the time component, but all choices can be explored.
Sure. Thanks
I think it is a combination of time series and geospatial
Good. Thanks for the info.
Looking at the date range in train and test shows that the submissions are not evaluated in future time but over the same period. Note that we don't know how the private data is split but I assume it's similar to the test data used for the public score.
Date time features are probably useful but that's different from modelling the problem as a time-series forecasting.
This competition is very similar to the AirQo competition that ended a couple of months ago. I advise you to read the top solutions to get some ideas.
Be careful, there is a huge difference between time series forecasting and time series regression according to sktime website. I think this problem is more about time series regression.