cestwc/census-income

Dataset Card for Census Income (Adult)

This dataset is a precise version of Adult or Census Income. This dataset from UCI somehow happens to occupy two links, but we checked and confirm that they are identical.

We used the following python script to create this Hugging Face dataset.

import pandas as pd
from datasets import Dataset, DatasetDict, Features, Value, ClassLabel

# URLs
url1 = "https://archive.ics.uci.edu/ml/machine-learning-databases/adult/adult.data"
url2 = "https://archive.ics.uci.edu/ml/machine-learning-databases/adult/adult.test"

# Column names
columns = [
    "age", "workclass", "fnlwgt", "education", "education-num", "marital-status",
    "occupation", "relationship", "race", "sex", "capital-gain", "capital-loss",
    "hours-per-week", "native-country", "income"
]


# Load datasets
df_train = pd.read_csv(url1, names=columns, skipinitialspace=True)
df_test = pd.read_csv(url2, names=columns, skipinitialspace=True, skiprows=1)

# Convert continuous columns to float
continuous_columns = ["age", "fnlwgt", "education-num", "capital-gain", "capital-loss", "hours-per-week"]
for col in continuous_columns:
    df_train[col] = pd.to_numeric(df_train[col], errors='coerce')
    df_test[col] = pd.to_numeric(df_test[col], errors='coerce')

df_test['income'] = df_test['income'].str.rstrip('.') # This is somewhat critical.

# Define categorical columns
categorical_columns = [
    "workclass", "education", "marital-status", "occupation", "relationship",
    "race", "sex", "native-country", "income"
]

# Dictionary to store category mappings
category_mappings = {}

for col in categorical_columns:
    # Convert train column to category and extract categories
    df_train[col] = df_train[col].astype("category")
    category_mappings[col] = df_train[col].cat.categories.to_list()  # Store category order

    # Apply the same category mapping to test
    df_test[col] = pd.Categorical(df_test[col], categories=category_mappings[col])

    # Convert to integer codes
    df_train[col] = df_train[col].cat.codes
    df_test[col] = df_test[col].cat.codes

# Define Hugging Face dataset schema
hf_features = Features({
    "age": Value("int64"),
    "workclass": ClassLabel(names=category_mappings["workclass"]),
    "fnlwgt": Value("int64"),
    "education": ClassLabel(names=category_mappings["education"]),
    "education-num": Value("int64"),
    "marital-status": ClassLabel(names=category_mappings["marital-status"]),
    "occupation": ClassLabel(names=category_mappings["occupation"]),
    "relationship": ClassLabel(names=category_mappings["relationship"]),
    "race": ClassLabel(names=category_mappings["race"]),
    "sex": ClassLabel(names=category_mappings["sex"]),
    "capital-gain": Value("int64"),
    "capital-loss": Value("int64"),
    "hours-per-week": Value("int64"),
    "native-country": ClassLabel(names=category_mappings["native-country"]),
    "income": ClassLabel(names=category_mappings["income"])
})

# Convert pandas DataFrame to Hugging Face Dataset
hf_train = Dataset.from_pandas(df_train, features=hf_features)
hf_test = Dataset.from_pandas(df_test, features=hf_features)

# Create a dataset dictionary
hf_dataset = DatasetDict({
    "train": hf_train,
    "test": hf_test
})

# Print dataset structure
print(hf_dataset)

The printed output could look like

DatasetDict({
    train: Dataset({
        features: ['age', 'workclass', 'fnlwgt', 'education', 'education-num', 'marital-status', 'occupation', 'relationship', 'race', 'sex', 'capital-gain', 'capital-loss', 'hours-per-week', 'native-country', 'income'],
        num_rows: 32561
    })
    test: Dataset({
        features: ['age', 'workclass', 'fnlwgt', 'education', 'education-num', 'marital-status', 'occupation', 'relationship', 'race', 'sex', 'capital-gain', 'capital-loss', 'hours-per-week', 'native-country', 'income'],
        num_rows: 16281
    })
})