census = pd.read_csv('hw5_census_dist/train_data.csv')
census = census.drop('fnlwgt', axis=1)
for category in census.keys():
    replace = None
    if type(census[category][0]) == str:
        replace = census[category].mode()[0]
    else:
        replace = int(census[category].mean())
    census[category] = census[category].replace('?', replace)
census = shuffle(census)
length = int(0.2 * len(census))
census_training_data = pd.DataFrame.as_matrix(census.drop('label', axis=1)[length:])
census_validation_data = pd.DataFrame.as_matrix(census.drop('label', axis=1)[:length])
census_training_labels = pd.DataFrame.as_matrix(census['label'][length:])
census_validation_labels = pd.DataFrame.as_matrix(census['label'][:length])