Refactor the package

[eduardocorrearaujo]

By now, the epigraphhub is organized as follows: ** - represents a file inside a subfolder Epigraphhub/

• analysis/:

  • clustering.py
  • epistats.py
  • restim.py

• data/

  • epigraphhub.py
  • foph.py
  • ggtrends.py
  • preprocessing.py
  • upload.py
  • worldbank.py
  • worldpop.py

• models /

  • lstm_models.py
  • ngboost_models.py
  • plots.py
  • sklearn_models.py

And the examples folder is organized as follows:

Examples/

• trends.ipynb;
• worlbank.ipynb;
• worldpop.ipynb.
• forecast_switzerland/

I propose the change:

Epigraphhub/

• analysis/

  • temporal/ ** clustering.py
  • epistats.py
  • restim.py

• data/

  • epigraphhub.py
  • covid/ ** foph.py
    • ggtrends.py
    • preprocessing.py (move to models folder)
    • upload.py (rename to upload_geofile.py)
      • indicators/ worldbank.py worldpop.py

• forecast_models (rename models) /

  • preprocessing.py
  • ml/ sklearn_models.py ngboost_models.py ** plots.py (maybe we could rename to make_plots.py
  • nn (neural network)/ ** lstm_models.py

And the examples folder it will be organized as follows:

The idea is the notebooks has the same name of the .py files and following the same structure shown above:

Examples/

• analysis
• data
• models
• forecast_switzerland/

[xmnlab]

I think maybe it would be better if we rename the Examples folder to tutorial, and move it to inside the docs folder. Also, I recommend to standardize the name of the folders to use always lower case.

[eduardocorrearaujo]

I agree, thank you for your contribution, @xmnlab.

[fccoelho]

Some initial suggestions:

• Move all plot related code in to a subpackage named plottingor something similar.
• in the data packages create subpackages by sources and inside we can have modules by dataset from that source.
• models should be a subpackage of analysis. Inside we can have modules or subpackages for types of models.

[eduardocorrearaujo]

ok, thank you

[eduardocorrearaujo]

Some initial suggestions:

• Move all plot related code in to a subpackage named plottingor something similar.
• in the data packages create subpackages by sources and inside we can have modules by dataset from that source.
• models should be a subpackage of analysis. Inside we can have modules or subpackages for types of models.

I was thinking about keeping the forecast models in a separate folder. Otherwise, I think that the path to the models will be too long: "from epigraphhub.analysis.models.forecast.ngboost_models import xxxxxxx" It's just an opinion looking at the scipy docs.

Other point that I saw in scipy, is keep the tests in the folder with the code instead of the main folder. What do you think about it?

[eduardocorrearaujo]

It was solved by PR #167