https://github.com/nixtla logo
#neural-forecast
Title
# neural-forecast
f

Fidele Koffivi Gbagbe

07/20/2023, 5:51 PM
Hello I'm using the library for the first time and I have an issue. I splited my data into train, val and test After that I fit the LSTM model "model.fit(train_data_)" . But I'm getting an error when trying to evaluate the model p = model.predict(futr_df=eval_data_).reset_index() Here is a pic of the error message
c

Cristian (Nixtla)

07/20/2023, 5:56 PM
Hi @Fidele Koffivi Gbagbe! The parameter
futr_df
is only to be used for the future exogenous covariates (see example in https://nixtla.github.io/neuralforecast/examples/exogenous_variables.html). You should use
df
to pass the usual dataset (with
y
and historic exogenous covariates), and the function will return the next
h
forecasts after the last date of
df
.
1
If you want to make rolled forecasts (for many timestamps in the
eval_data
), you can merge
train_data
and
eval_data
and use the
cross_validation
method.