How forecast values using ARIMAX model in Python?

Question

I am trying to fit ARIMAX model on train sample (endogenous and exogenous variables) and then forecast using exogenous variables (they are available). I am using statsmodels module in Python.

I have the following code:

#split datasets
df_train = df.iloc[:100]
df_test= df.iloc[100:104]

# Define the model
model = ARIMA(endog= df_train['y'], exog=df_train[['x1', 'x2']], order=(2,0,2))
# Fit the model
results = model.fit()

#predict for the next 5 periods
results.predict(steps = 5,  exog = df_test[['pc1', 'pc2']])

Unfortunately, seems it predicts in-sample fit using train dataset but not test dataset, because there are 100 prediction points.

If there is 2 lags in the model, so should I append last 2 points of y from train dataset or should not ( results somehow preserves information about last value of y )?

Similar question I already found, however, they were related to R.

Answer 1

Use the forecast method:

results.forecast(steps=5,  exog=df_test[['pc1', 'pc2']])