Function apply_transforms

alphapy/features.py:155–223

r"""Apply special functions to the original features. Parameters ---------- model : alphapy.Model Model specifications indicating any transforms. X : pandas.DataFrame Combined train and test data, or just prediction data. Returns ------- all_features : p

(model, X)

#

def apply_transforms(model, X):
    r"""Apply special functions to the original features.

    Parameters
    ----------
    model : alphapy.Model
        Model specifications indicating any transforms.
    X : pandas.DataFrame
        Combined train and test data, or just prediction data.

    Returns
    -------
    all_features : pandas.DataFrame
        All features, including transforms.

    Raises
    ------
    IndexError
        The number of transform rows must match the number of
        rows in ``X``.

    """

    # Extract model parameters
    transforms = model.specs['transforms']

    # Log input parameters

    logger.info("Original Features : %s", X.columns)
    logger.info("Feature Count     : %d", X.shape[1])

    # Iterate through columns, dispatching and transforming each feature.

    logger.info("Applying transforms")
    all_features = X

    if transforms:
        for fname in transforms:
            # find feature series
            fcols = []
            for col in X.columns:
                if col.split(LOFF)[0] == fname:
                    fcols.append(col)
            # get lag values
            lag_values = []
            for item in fcols:
                _, _, _, lag = vparse(item)
                lag_values.append(lag)
            # apply transform to the most recent value
            if lag_values:
                f_latest = fcols[lag_values.index(min(lag_values))]
                features = apply_transform(f_latest, X, transforms[fname])
                if features is not None:
                    if features.shape[0] == X.shape[0]:
                        all_features = pd.concat([all_features, features], axis=1)
                    else:
                        raise IndexError("The number of transform rows [%d] must match X [%d]" %
                                         (features.shape[0], X.shape[0]))