While providing standard models for forecasting is important to make Haskell a viable data science platform -- most data scientists I know just use 'ready-to-use' models -- I think we can accelerate development by providing a nice interface to specify arbitrary models, and implement the popular models in this interface
A common use of (time-) series is for forecasting. Popular methods include:
Peek at Nixtla's statsforecast models to get a sense of standard forecasting methods.
While providing standard models for forecasting is important to make Haskell a viable data science platform -- most data scientists I know just use 'ready-to-use' models -- I think we can accelerate development by providing a nice interface to specify arbitrary models, and implement the popular models in this interface