TabulaCopula

class TabulaCopula(definitions=None, output_general_prefix=None, conditionalSettings_dict=None, metaData_transformer=None, var_list_filter=None, removeNull=False, sampling=None, debug=False) Module for performing copula/conditional-copula (Gaussian) for Tabular-type data.

Parameters

definitions: file (.py), optional, default None. Contain global variables

output_general_prefix: str, optional, default None. Prefix used for all output files, e.g. "EXPT_1". If not None, replaces settings in definitions.

conditionalSettings_dict: dict, optional, default None. Dictionary of conditional inputs for using conditional-copula.

metaData_transformer: dict, optional, default None. Dictionary of inputs for the Transformer class initialisation (ref Transformer metaData).

var_list_filter: list, optional, default None. List of variables to transform. If None, all will be transformed.

removeNull: boolean, optional, default False. Whether to remove all null values prior to transformation.

sampling: float, optional, default None. Percentage of sample points draw from the transformed dataframe, leaving the rest as control. If None, all training points will be used. (Note that the sampling process is done after the transformation.)

debug: boolean, default True. Whether to print debug-related outputs to console.

Notes

Description of conditionalSettings_dict

The conditionalSettings_dict variable specifies the structure of the conditional setup. It takes the following form:

conditionalSettings_dict = {
    "set_1": {
        "bool": True,
        "parent_conditions": { # parents, the `Y` in `P(X | Y)` while learning P(X | Y).
            "SurveyYr": { # split variable into 2 sets
                "condition": "set", #available options: "set", "range"
                "condition_value": {
                    1: ["2009_10"],
                    2: ["2011_12"]
                }
            },
            "Age": { # split variable into 3 sets based on range
                "condition": "range",
                "condition_value": {
                    1: [">=3", "<79"],
                    2: ["<3"],
                    3: [">=79"]
                }
            }
        },
        "conditions_var": ["Age"], # the `Y` to keep constant while generating values of `X` in `P(X | Y)`. Can be a float, in which case it is a threshold to fix all variables with pairwise correlation (with X) above then said threshold.
        "children": ['AgeMonths'] #variable for which to learn the joint conditional distributions on, the `X` in `P(X | Y)`. Can be a string: "allOthers".
    }
}

Examples

Please refer to the below pages for detailed examples:

Attributes

Methods

Table of contents