hello I don't understand why these both target transformations (Autodifference and AutoSeasonalityAndDifferences ) don't give the same result when the unique id has only a trend ? Does it mean that it is necessary to isolate products that have seasonality before to transform by AutoSeasonalityAndDifferences ? Thank you in advance for your help

Hey. The AutoSeasonalityAndDifferences determines the season length (up to the max_season_length and takes that difference if it's significant) but doesn't consider single diferences, so you'd need to combine both to remove trend and seasonalities

Thank you for your answer but i am bit confused because if I take an other example with a strong seasonality and a trend, then AutoSeasonalityAndDifferences(max_season_length=12, max_diffs=1) removed the seasonality AND trend :s

Does that look the same if you take the 12th difference?

AutoDifferences(max_diffs=1)

I meant just Differences([12])

Differences([12])

Seems like it's removing both since the trend is additive and matches exactly the seasonal period

i have added noise to not have perfect data : AutoSeasonalityAndDifferences(max_season_length=12, max_diffs=1) removed the trend and the seasonality

It removes the seasonality, which can also remove the trend if it matches. If you want to remove both for sure you can use AutoDifferences and then AutoSeasonalityAndDifferences

ok fine. thank you.