adult.py

from typing import List
from functools import partial

import datasets

import pandas


VERSION = datasets.Version("1.0.0")
_ORIGINAL_FEATURE_NAMES = [
    "age",
    "workclass",
    "final_weight",
    "education",
    "education-num",
    "marital_status",
    "occupation",
    "relationship",
    "race",
    "sex",
    "capital_gain",
    "capital_loss",
    "hours_per_week",
    "native_country",
    "threshold"
]
_BASE_FEATURE_NAMES = [
    "age",
    "capital_gain",
    "capital_loss",
    "education",
    "final_weight",
    "hours_worked_per_week",
    "marital_status",
    "native_country",
    "occupation",
    "race",
    "relationship",
    "is_male",
    "workclass",
    "over_threshold",
]
_ENCODING_DICS = {
    "is_male": {
            "Male": True,
            "Female": False
        }
}
_RACE_ENCODING = {
    "White": 0,
    "Black": 1,
    "Asian-Pac-Islander": 2,
    "Amer-Indian-Eskimo": 3,
    "Other": 4,
}
_EDUCATION_ENCODING = {
	"Preschool": 0,
	"1st-4th": 1,
	"5th-6th": 2,
	"7th-8th": 3,
	"9th": 4,
	"10th": 5,
	"11th": 6,
	"12th": 7,
	"HS-grad": 8,
	"Assoc-acdm": 9,
	"Assoc-voc": 10,
	"Some-college": 11,
    "Bachelors": 12,
	"Masters": 13,
	"Doctorate": 14,
	"Prof-school": 15
}

DESCRIPTION = "Adult dataset from the UCI ML repository."
_HOMEPAGE = "https://archive.ics.uci.edu/ml/datasets/Adult"
_URLS = ("https://huggingface.co/datasets/mstz/adult/raw/adult.csv")
_CITATION = """
@inproceedings{DBLP:conf/kdd/Kohavi96,
  author    = {Ron Kohavi},
  editor    = {Evangelos Simoudis and
               Jiawei Han and
               Usama M. Fayyad},
  title     = {Scaling Up the Accuracy of Naive-Bayes Classifiers: {A} Decision-Tree
               Hybrid},
  booktitle = {Proceedings of the Second International Conference on Knowledge Discovery
               and Data Mining (KDD-96), Portland, Oregon, {USA}},
  pages     = {202--207},
  publisher = {{AAAI} Press},
  year      = {1996},
  url       = {http://www.aaai.org/Library/KDD/1996/kdd96-033.php},
  timestamp = {Mon, 05 Jun 2017 13:20:21 +0200},
  biburl    = {https://dblp.org/rec/conf/kdd/Kohavi96.bib},
  bibsource = {dblp computer science bibliography, https://dblp.org}
}"""

# Dataset info
urls_per_split = {
    "train": "https://huggingface.co/datasets/mstz/adult/raw/main/adult_tr.csv", 
    "test": "https://huggingface.co/datasets/mstz/adult/raw/main/adult_ts.csv"
}
features_types_per_config = {
    "encoding": {
        "feature": datasets.Value("string"),
        "original_value": datasets.Value("string"),
        "encoded_value": datasets.Value("int64"),
    },
    "income": {
        "age": datasets.Value("int64"),
        "capital_gain": datasets.Value("float64"),
        "capital_loss": datasets.Value("float64"),
        "education": datasets.Value("int8"),
        "final_weight": datasets.Value("int64"),
        "hours_worked_per_week": datasets.Value("int64"),
        "marital_status": datasets.Value("string"),
        "native_country": datasets.Value("string"),
        "occupation": datasets.Value("string"),
        "race": datasets.Value("string"),
        "relationship": datasets.Value("string"),
        "is_male": datasets.Value("bool"),
        "workclass": datasets.Value("string"),
        "over_threshold": datasets.ClassLabel(num_classes=2, names=("no", "yes"))
    },
    "income-no race": {
        "age": datasets.Value("int64"),
        "capital_gain": datasets.Value("float64"),
        "capital_loss": datasets.Value("float64"),
        "education": datasets.Value("int64"),
        "final_weight": datasets.Value("int64"),
        "hours_worked_per_week": datasets.Value("int64"),
        "marital_status": datasets.Value("string"),
        "native_country": datasets.Value("string"),
        "occupation": datasets.Value("string"),
        "relationship": datasets.Value("string"),
        "is_male": datasets.Value("bool"),
        "workclass": datasets.Value("string"),
        "over_threshold": datasets.ClassLabel(num_classes=2, names=("no", "yes"))
    },
    "race": {
        "age": datasets.Value("int64"),
        "capital_gain": datasets.Value("float64"),
        "capital_loss": datasets.Value("float64"),
        "education": datasets.Value("int64"),
        "final_weight": datasets.Value("int64"),
        "hours_worked_per_week": datasets.Value("int64"),
        "marital_status": datasets.Value("string"),
        "native_country": datasets.Value("string"),
        "occupation": datasets.Value("string"),
        "relationship": datasets.Value("string"),
        "is_male": datasets.Value("bool"),
        "workclass": datasets.Value("string"),
        "over_threshold": datasets.Value("int8"),
        "race": datasets.ClassLabel(num_classes=5, names=["White", "Black", "Asian-Pac-Islander", "Amer-Indian-Eskimo", "Other"])
    }
}
features_per_config = {k: datasets.Features(features_types_per_config[k]) for k in features_types_per_config}


class AdultConfig(datasets.BuilderConfig):
    def __init__(self, **kwargs):
        super(AdultConfig, self).__init__(version=VERSION, **kwargs)
        self.features = features_per_config[kwargs["name"]]


class Adult(datasets.GeneratorBasedBuilder):
    # dataset versions
    DEFAULT_CONFIG = "income"
    BUILDER_CONFIGS = [
        AdultConfig(name="income",
                    description="Adult for income threshold binary classification."),
        AdultConfig(name="income-no race",
                    description="Adult for income threshold binary classification, race excluded from features."),
        AdultConfig(name="race", 
                    description="Adult for race (multiclass) classification."),
        AdultConfig(name="encoding",
                    description="Encoding dictionaries for discrete features.")
    ]


    def _info(self):       
        info = datasets.DatasetInfo(description=DESCRIPTION, citation=_CITATION, homepage=_HOMEPAGE,
                                    features=features_per_config[self.config.name])

        return info
    
    def _split_generators(self, dl_manager: datasets.DownloadManager) -> List[datasets.SplitGenerator]:
        downloads = dl_manager.download_and_extract(urls_per_split)

        return [
            datasets.SplitGenerator(name=datasets.Split.TRAIN, gen_kwargs={"filepath": downloads["train"]}),
            datasets.SplitGenerator(name=datasets.Split.TEST, gen_kwargs={"filepath": downloads["test"]}),
        ]
    
    def _generate_examples(self, filepath: str):
        if self.config.name == "encoding":
            data = self.encodings()

            for row_id, row in data.iterrows():
                data_row = dict(row)

                yield row_id, data_row

        else:
            data = pandas.read_csv(filepath)
            data = self.preprocess(data, config=self.config.name)

            for row_id, row in data.iterrows():
                data_row = dict(row)

                yield row_id, data_row

    def encodings(self):
        data = pandas.DataFrame([("education", original_value, encoded_value)
                                 for original_value, encoded_value in _EDUCATION_ENCODING.items()],
                                columns=["feature", "original_value", "encoded_value"])

        return data


    def preprocess(self, data: pandas.DataFrame, config: str = DEFAULT_CONFIG) -> pandas.DataFrame:
        data.drop("education", axis="columns", inplace=True)
        data = data.rename(columns={"threshold": "over_threshold", "sex": "is_male"})

        data = data[["age", "capital_gain", "capital_loss", "education-num", "final_weight",
                     "hours_per_week", "marital_status", "native_country", "occupation",
                     "race", "relationship", "is_male", "workclass", "over_threshold"]]
        data.columns = _BASE_FEATURE_NAMES

        for feature in _ENCODING_DICS:
            encoding_function = partial(self.encode, feature)
            data.loc[:, feature] = data[feature].apply(encoding_function)
        

        if config == "income":
            return data[list(features_types_per_config["income"].keys())]
        elif config == "income-no race":
            return data[list(features_types_per_config["income-no race"].keys())]
        elif config =="race":
            data.loc[:, "race"] = data.race.apply(self.encode_race)
            data = data[list(features_types_per_config["race"].keys())]

            return data
    
    def encode(self, feature, value):
        if feature in _ENCODING_DICS:
            return _ENCODING_DICS[feature][value]
        raise ValueError(f"Unknown feature: {feature}")
    
    def encode_race(self, race):
        return _RACE_ENCODING[race]