Hello! Does anyone has any guidance on how to use ...
# neural-forecasty
Yeray Pabon
03/12/2024, 2:02 PMHello! Does anyone has any guidance on how to use the future exogenous variables in the case of cross validation? Usually. I use
Copy code
Y_hat_df = nf.predict(futr_df=futr_df)Yeray Pabon
03/12/2024, 2:18 PMI would like to also put a second question here. Let's say there'll be a future event on a specific hour that will affect the forecasting accuracy. How it's best recommended to format this futr_df? Would it be like an array with 1s and 0s where 1 = Event, and 0 = No Event?
c
Cristian (Nixtla)
03/12/2024, 2:25 PMHi @Yeray Pabon. The cross-validation method already takes cares of using the appropriate future variables for each window. The input dataframe needs to have all historic and future variables in columns, and the method will handle them differently. Note that you dont need to shift any of them, just include the corresponding value to each timestamp
Cristian (Nixtla)
03/12/2024, 2:26 PMYes, the event dummy is a common practice. You can also create a distance to/from event variable
y
Yeray Pabon
03/12/2024, 2:29 PMAwesome! Didn't know the method already takes care of it. Thank you, @Cristian (Nixtla)!