pythonarimapmdarima

pmdarima save auto.arima aic values

Is there a way to save this AIC values?

I want to sort this values to see which model has lowest AIC in order to make cross validation to the ones with lower AIC'senter image description here

Solution

  • I saw this issue in the GitHub page of pmd arima (alkaline-ml/pmdarima).

    The answer is from Taylor G. Smith, who actively maintain pmdarima.

    There is two ways.

    1)

    The first one is with the argument return_valid_fits set as True. See pmdarima 1.8.0 documentation:

    return_valid_fits : bool, optional (default=False)

    If True, will return all valid ARIMA fits in a list. If False (by default), will only return the best fit.

    Example:

    import pmdarima as pm

sxmodel = pm.auto_arima(endog[:n_train],exog[:n_train], start_p=0, start_q=0, max_p=2, max_q=2,
                           start_P=0,start_Q=0, max_P=2,max_D=1,max_Q=2, m=7, seasonal=True,
                           d=0, trace=True, error_action='trace',suppress_warnings=True, stepwise=True)

sxmodel

    In this first case the sxmodel will be a tuple with information about the fitted models

    2)

    The other one is using the sys module:

    import sys
orig_stdout = sys.stdout
f = open('out.txt', 'w')
sys.stdout = f

# fit your model
model = pm.auto_arima(...)

sys.stdout = orig_stdout
f.close()