Here's an example of Cross-validation for PyFlux - needs some refactoring

cross_validation.py

def time_series_cross_validation_2(model, data, params, number_folds=10):
    """
        Performs a k-fold cross-validation on a Time Series as described by
        Rob Hyndman.
        See Also:
            http://robjhyndman.com/hyndsight/crossvalidation/
        Arguments: 
            model: a model somewhere... 
            database (numpy.ndarray): uses 'data' matrix to perform
                cross-validation.
            params (dict): dictionary of parameters to train/test.
            number_folds (int): number of folds to be created from training and
                testing matrices.
        Returns:
            tuple: tuple of AIC between training and testing for out of sample
    """

    if number_folds < 2:
        print("Error: Must have at least 2-folds.")
        return

    number_patterns = data.shape[0]
    fold_size = round(number_patterns / number_folds)

    folds = []
    for k in range(number_folds):
        folds.append(database[k * fold_size:(k + 1) * fold_size])

    training_errors = []
    testing_errors = []

    training_matrix = folds[0]
    testing_matrix = []
    for k in range(number_folds - 1):
        if k > 0:
            training_matrix = \
                np.concatenate((training_matrix, testing_matrix), axis=0)

        testing_matrix = folds[k + 1]
        print("Testing matrix:", testing_matrix)

        tr_error = model(training_matrix, **params)
        model_tr = tr_error.fit('BBVI',iterations=1000).aic
        te_error = tr_error.predict(testing_matrix.shape[0]).values
        print("Rmse error", rmse(te_error[0], testing_matrix))


    return print("Training AIC: ", training_errors)