UCI Datasets

quack bundles loaders for ~30 datasets from the UCI Machine Learning Repository, following the same selection used in [1]. Each dataset is wrapped by a BaseDatasetLoader subclass, registered under a short string key in UCILoaderFactory, and exposed through the single load_uci function.

Quickstart

from quack.datasets import load_uci

X, y = load_uci("bc-cont")

print(X.shape, y.shape)

load_uci downloads the raw source file(s) (cached via urllib), applies the dataset-specific preprocessing (encoding categoricals, binning continuous targets, dropping constant/leaky columns, etc.), and returns X as float32 and y as the label array — ready to be split and fed into any quantifier's .fit(X, y).

Listing available datasets

from quack.datasets import UCI_DATASETS

print(UCI_DATASETS)
# ['adult', 'avila', 'bike', 'blog', 'bc-cont', 'bc-int', 'cars', 'conc',
#  'contra', 'cappl', 'ccard', 'dota', 'drug', 'ener', 'flare', 'grid',
#  'ads', 'magic', 'boone', 'mush', 'music', 'news', 'nurse', 'occup',
#  'spam', 'cond', 'turk', 'wine', 'yeast']

Dataset reference

Key	Dataset	Task type / target binning
`adult`	Adult (Census Income)	Binary (`<=50K` / `>50K`)
`avila`	Avila	Binary (class A vs. rest)
`bike`	Bike Sharing	4-class (`cnt` binned)
`blog`	BlogFeedback	4-class (`att280` binned)
`bc-cont`	Breast Cancer Wisconsin (Original)	Binary (`diagnosis`)
`bc-int`	Breast Cancer Wisconsin (Diagnostic)	Binary (`Class`)
`cars`	Car Evaluation	4-class
`conc`	Concrete Compressive Strength	3-class (`strength` binned)
`contra`	Contraceptive Method Choice	3-class
`cappl`	Credit Approval	Binary
`ccard`	Default of Credit Card Clients	Binary
`dota`	Dota2 Games Results	Binary (`Winner`)
`drug`	Drug Consumption	3-class (grouped `CL0-CL6`)
`ener`	Appliances Energy Prediction	3-class (`Appliances` binned)
`flare`	Solar Flare	Binary (`C` binned)
`grid`	Electrical Grid Stability Simulated	Binary (`stabf`)
`ads`	Internet Advertisements	Binary (`class`)
`magic`	MAGIC Gamma Telescope	Binary (`target`)
`boone`	MiniBooNE Particle Identification	Binary (`signal`)
`mush`	Mushroom	Binary (`result`)
`music`	Geographical Original of Music	Binary (`att117` binned)
`news`	News Popularity	4-class (`shares` binned)
`nurse`	Nursery	3-class
`occup`	Occupancy Detection	Binary (`Occupancy`)
`spam`	Spambase	Binary (`spam`)
`cond`	Superconductivity	4-class (`critical_temp` binned)
`turk`	Turkiye Student Evaluation	Multi-class (`instr`)
`wine`	Wine Quality (red + white)	4-class (`quality` binned)
`yeast`	Yeast	4-class (grouped)

Info

Some loaders (ccard, dota, boone, avila, music, bike, blog, ener, news, cond) download .zip/.xls archives on demand and may take a while on first use, since no local cache directory is used by these UCI loaders (unlike load_forman, which caches under ~/.quack_data/).

Combining with a quantifier

from sklearn.model_selection import train_test_split
from sklearn.linear_model import LogisticRegression
from quack.datasets import load_uci
from quack.quantifiers import ACC
from quack.metrics import ae
import numpy as np

X, y = load_uci("magic")
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3, random_state=42)

quantifier = ACC(classifier=LogisticRegression(max_iter=1000))
quantifier.fit(X_train, y_train)

labels, counts = np.unique(y_test, return_counts=True)
true_prev = counts / y_test.shape[0]
pred_prev = quantifier.predict(X_test)

print(f"AE: {ae(true_prev, pred_prev):.4f}")

References

[1] Schumacher, T., Strohmaier, M., & Lemmerich, F. (2025). A comparative evaluation of quantification methods. Journal of Machine Learning Research, 26(55), 1-54.