label leakage datasets

Here's an example "bot" that detects label leakage:

from sklearn.tree import DecisionTreeClassifier
from sklearn.preprocessing import Imputer

import openml

# get all classification tasks
tasks = openml.tasks.list_tasks(task_type_id=1)

# get all "small" datasets that are used for classification
dids = set()
for tid, task in tasks.items():
    if task['NumberOfFeatures'] * task['NumberOfInstances'] < 10000:
        dids.add(task['did'])

for did in dids:
    try:
        ds = openml.datasets.get_dataset(did)
        X, y = ds.get_data(target=ds.default_target_attribute)

    except:
        print("error")
    X = Imputer().fit_transform(X)
    n_classes = int(float(ds.qualities['NumberOfClasses']))
    if n_classes == -1:
        continue
    acc = DecisionTreeClassifier(max_leaf_nodes=n_classes).fit(X, y).score(X, y)
    if acc > .99:
        print(did, acc)

results:
480 1.0
694 1.0
759 1.0
778 1.0
818 1.0
874 1.0
880 1.0
885 1.0
951 1.0
965 1.0
969 1.0
1121 1.0
1167 1.0

The issue names no repository files or tests. Start with the provided Python example and dataset IDs 480, 694, 759, 778, 818, 874, 880, 885, 951, 965, 969, 1121, and 1167; inspect their metadata and feature/target relationships, then clarify whether done means detection, correction, or documentation.