This message was deleted.
# generals
Slackbot
11/22/2022, 1:26 PMThis message was deleted.
m
Max (Nixtla)
11/22/2022, 3:02 PMHi! For production settings we recommend using the forecast method instead of the fit predict ones.
Max (Nixtla)
11/22/2022, 3:03 PMAlso, you should explore the possibility of using other models like MSTL or ETS.
This newly released guide might be helpful.
https://nixtla.github.io/statsforecast/examples/getting_started_complete.html
a
Andrei Tulbure
11/23/2022, 7:55 AM@Max (Nixtla) https://github.com/Nixtla/statsforecast/tree/main/experiments/arima_prophet_adapter -> I see you here using fit and then make future dataframe and then predict. Could you be more specific what you mean by using forecast?
Andrei Tulbure
11/23/2022, 7:56 AMSecondly, we want to test, on our dataset, how much of a speed and accuracy improvement we get
Andrei Tulbure
11/23/2022, 5:09 PM?
m
Max (Nixtla)
11/23/2022, 5:09 PMHi Andrei, I will try to give you a complete answer soon.
✅ 1
Max (Nixtla)
11/23/2022, 5:09 PMDo you have any code that you could share so i can be more specific?
a
Andrei Tulbure
11/23/2022, 5:11 PMYes, I will ask the guys from our team to send over the code. But most plainly put. compared directly Autoarimaprophet with prophet, with uncert intervals = 0 and some fut regressors. The dataset is weekly and kinda small.
m
Max (Nixtla)
11/23/2022, 5:13 PMIs there a specific reason you want to compare Prophet and AutoArima? I would highly suggest finding appropriate models for each combination of series (e.g. product and store).
For some series, sparse models like ADIDA or Croston might be very successful.
a
Andrei Tulbure
11/23/2022, 5:14 PMwe have weekly data and it`s not sparse. And we want to use it because 1. we have seasonality and trend + maybe some holidy effect (product demand forecasting). we have some future regressors and dataset is not that large to use deep learning models
m
Max (Nixtla)
11/23/2022, 5:20 PMFor that use case, I suggest comparing Prophet against MSTL.
MSTL decomposes seasonality and trends but can fit multiple seasonalities. It’s also faster than ARIMA.
That’s good for your use case because with weekly data, you might find monthly patterns (last week of the month) as well as yearly patterns (last week of December)
✅ 1
Max (Nixtla)
11/23/2022, 5:21 PMWould you mind looking at this guide and telling me if that’s something you would try?
✅ 1
8 Views