object_analyzer.py 文件源码

def analyze_pd_dataframe(dataframe, target_attributes):
    """Analyze pandas.Dataframe and convert it into internal representation.

    Parameters
    ----------
    dataframe : pd.Dataframe
        input data, can contain float, int, object

    target_attributes : int, str or list
        Index the target attribute. If this is
        * an int, use this as an index (only works with positive indices)
        * a str, use this to compare with the column values
        * a list (which must either consist of all ints or strs), of which
          all elements that matched are assumed to be targets.

    Returns
    -------
    np.ndarray
        Data. All columns are converted to type float. Categorical data is
        encoded by positive integers.

    dict
        Attribute types. Contains the following keys:
        * `type`: `categorical` or 'numerical`
        * `name`: column name of the dataframe
        * `is_target`: whether this column was designated as a target column

    """
    dataframe = _normalize_pd_column_names(dataframe)
    attribute_types = _get_pd_attribute_types(dataframe, target_attributes)
    dataframe = _replace_objects_by_integers(dataframe, attribute_types)

    return dataframe.values, attribute_types