There are a lot of inconsistencies between the packages that make using the nixtla "universe" onerous to use. Here a few I've noticed: •

statsforecast

use

h

as horizon, while

mlforecast

uses

horizon

for the same parameter. •

statsforecast

has both

forecast

and

forecast_fitted_values

functions, while

mlforecast

does not. •

statsforecast

has a parameter

n_jobs

, while

mlforecast

has

num_threads

. •

unique_id

as the data index.

Thanks! That is a very good observation and we are working on making everything more homogeneous. cc @José Morales @fede (nixtla) (they/them)

Hey @Isaac,

unique_id

doesn't have to be index in statsforecast,

DataFrameProcessing

extracts

unique_id

if it is index or not, either way is fine.