Apply a prophet model to new data without re-training the model

[Yasslight90]

Hi there,

I was wondering if there is a way to apply a prophet model (already learned) to a new data. Let say, I learned a model M from a time series ($y_1$, ..., $y_T$) and used it to predict the next value $y_{T+1}$. What if I want to predict $y_{T+2}$ based on ($y_1$, ..., $y_T$, $y_{T+1}$) without relearning the model? is it possible with Prophet ?

With ARIMA, this is possible using the function apply

Thank you!

[ghost]

Prophet is a GAM model that attempts a curve fitting exercise. Given this, it is recommended to always retrain the model. Unlike with an ARIMA model where the future values would be predicted by the model and then those predicted values are used to predict future values, Prophet does not do this so to take into consideration local changes in trend, seasonality, etc. it is recommended to retrain.

If you would like to use a previously saved model to predict on new data, you can use can refer to the documentation on this in terms of saving a model or updating fitted models.