SNU_gen.py

'''Genreates a syntethic dataset (csv) of persons to test the SNU_assignator

Usage:
    SNU_gen.py <o> [options]

Arguments:
    <o>                   An output path to store the ysntethic data csv
    -n PER                   Number of persons to generate [default: 2000:int]
    -f FIL                  Representation proportion of the filiere. Ex: "0.1,0.1,...,0.1" (default: None)
    -r RES               Representation proportion of the residence Ex: "0.1,0.1,...,0.1" (default: None)
    -s SEX               Representation proportion of the residence Ex: "0.3,0.7" (default: None)
'''

import logging
from math import isclose
import pandas as pd

import numpy as np
from argopt import argopt

logger = logging.getLogger()
logger.setLevel(logging.INFO)

FIXED_VALUES = {
    "filiere": ["LGT", "LP", "MILO", "CFA", "EMS", "PJJ", "SHN", "Actif"],
    "residence": ["Ardennes", "Puy-de-Dôme", "Cher", "Haute-Saône", "Morbihan", "Eure", "Haute-Pyrénées",
                  "Loire-Atlantique",
                  "Vaucluse", "Guyane", "Nord", "Creuse", "Val d'Oise"],
    "sexe": ["F", "M"]
}


def generate_syn_data(output_path, n_persons, proportions):
    dict_values = {}
    for col in ["filiere", "residence", "sexe"]:
        dict_values[col] = np.random.choice(FIXED_VALUES[col], n_persons, p=proportions[col])

    df = pd.DataFrame(dict_values, index=np.arange(n_persons))
    df.to_csv(output_path, encoding="utf8")
    pass


if __name__ == '__main__':
    parser = argopt(__doc__).parse_args()
    output_path = parser.o
    n_persons = parser.n
    proportions = {}

    proportions["filiere"] = parser.f
    proportions["residence"] = parser.r
    proportions["sexe"] = parser.s

    for value, proportion in proportions.items():
        # TODO check if any of these values asigned it is None
        if proportion:
            filiere_prop = np.array(list(map(float, proportion.split(","))))
            if len(filiere_prop) != len(FIXED_VALUES[value]) or not isclose(sum(filiere_prop), 1.):
                logger.error(
                    "{0} proportion was indicated but does not match the nummber of {1} available.".format(value, value),
                    "Or it is not a valid prob distribution (!= 1.0)",
                    "Using the same proportion for each one.")
                proportions[value] = np.ones(len(FIXED_VALUES[value])) / len(FIXED_VALUES[value])
            else:
                proportions[value] = filiere_prop
        else:
            proportions[value] = np.ones(len(FIXED_VALUES[value])) / len(FIXED_VALUES[value])
    generate_syn_data(output_path=output_path, n_persons=n_persons, proportions=proportions)