Seasonality modularization

[karl-richter]

🔬 Background

• Following the example of trend, the seasonality component is modularized.

🔮 Key changes

• Created a new Seasonality abstract and FourierSeasonality class.
• Find the new class diagram here.

📋 Review Checklist

• I have performed a self-review of my own code.
• I have commented my code, added docstrings and data types to function definitions.
• I have added pytests to check whether my feature / fix works.

Please make sure to follow our best practices in the Contributing guidelines.

[github-actions]

Model Benchmark

Benchmark	Metric	main	current	diff
PeytonManning	MAE_val	0.64636	0.64636	-0.0%	✅
PeytonManning	RMSE_val	0.79276	0.79276	-0.0%	✅
PeytonManning	Loss_val	0.01494	0.01494	-0.0%	✅
PeytonManning	MAE	0.42701	0.42701	0.0%	✅
PeytonManning	RMSE	0.57032	0.57032	-0.0%	✅
PeytonManning	Loss	0.00635	0.00635	0.0%	✅
PeytonManning	time	15.8239	20.6	30.18%	❌
YosemiteTemps	MAE_val	1.72949	1.72948	-0.0%	✅
YosemiteTemps	RMSE_val	2.27386	2.27386	-0.0%	✅
YosemiteTemps	Loss_val	0.00096	0.00096	-0.0%	✅
YosemiteTemps	MAE	1.45189	1.45189	-0.0%	✅
YosemiteTemps	RMSE	2.16631	2.16631	-0.0%	✅
YosemiteTemps	Loss	0.00066	0.00066	0.0%	✅
YosemiteTemps	time	124.291	169.76	36.58%	❌
AirPassengers	MAE_val	15.4077	15.4077	0.0%	✅
AirPassengers	RMSE_val	19.5099	19.5099	0.0%	✅
AirPassengers	Loss_val	0.00196	0.00196	0.0%	✅
AirPassengers	MAE	9.86947	9.86947	-0.0%	✅
AirPassengers	RMSE	11.7222	11.7222	-0.0%	✅
AirPassengers	Loss	0.00057	0.00057	-0.0%	✅
AirPassengers	time	5.48767	7.38	34.48%	❌

Model training plots

Model Training

PeytonManning

YosemiteTemps

AirPassengers

[alfonsogarciadecorral]

Hey @karl-richter
I'd suggest to follow the trend structure and creating Local/Global subclasses from FourierSeasonality.
It might not be that useful now, but it might be once we add Glocal.

[codecov-commenter]

Codecov Report

Merging #1141 (e755683) into main (55c23bb) will increase coverage by 0.06%.
The diff coverage is 93.65%.

📣 This organization is not using Codecov’s GitHub App Integration. We recommend you install it so Codecov can continue to function properly for your repositories. Learn more

@@            Coverage Diff             @@
##             main    #1141      +/-   ##
==========================================
+ Coverage   90.17%   90.24%   +0.06%     
==========================================
  Files          30       33       +3     
  Lines        4966     5000      +34     
==========================================
+ Hits         4478     4512      +34     
  Misses        488      488              

Impacted Files	Coverage Δ
neuralprophet/forecaster.py	87.90% <ø> (ø)
neuralprophet/hdays.py	99.73% <ø> (ø)
neuralprophet/plot_forecast_plotly.py	87.72% <ø> (ø)
neuralprophet/plot_model_parameters_plotly.py	94.89% <ø> (ø)
neuralprophet/time_net.py	90.16% <77.77%> (+0.05%)	⬆️
neuralprophet/components/router.py	87.50% <87.50%> (ø)
neuralprophet/components/seasonality/fourier.py	97.14% <97.14%> (ø)
neuralprophet/components/seasonality/__init__.py	100.00% <100.00%> (ø)
...euralprophet/components/seasonality/seasonality.py	100.00% <100.00%> (ø)
neuralprophet/plot_utils.py	89.82% <100.00%> (ø)
... and 2 more

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

[karl-richter]

Hey @karl-richter I'd suggest to follow the trend structure and creating Local/Global subclasses from FourierSeasonality. It might not be that useful now, but it might be once we add Glocal.

Good call @alfonsogarciadecorral , I adressed your comment and create a Global and Local class of Seasonality.

[ourownstory]

In a future PR, we may want to have a seasonality object per period instantiated - thus a class should not need know the num_seasonalities_modelled.

[ourownstory]

Great work!