"""Data check that checks if there are any columns in the input that are either too unique for classification problems or not unique enough for regression problems."""
from evalml.data_checks import (
DataCheck,
DataCheckActionCode,
DataCheckActionOption,
DataCheckMessageCode,
DataCheckWarning,
)
from evalml.problem_types import handle_problem_types, is_multiclass, is_regression
from evalml.utils.woodwork_utils import infer_feature_types
warning_not_unique_enough = (
"Input columns {} for {} problem type are not unique enough."
)
warning_too_unique = "Input columns {} for {} problem type are too unique."
[docs]class UniquenessDataCheck(DataCheck):
"""Check if there are any columns in the input that are either too unique for classification problems or not unique enough for regression problems.
Args:
problem_type (str or ProblemTypes): The specific problem type to data check for.
e.g. 'binary', 'multiclass', 'regression, 'time series regression'
threshold(float): The threshold to set as an upper bound on uniqueness for classification type problems
or lower bound on for regression type problems. Defaults to 0.50.
"""
def __init__(self, problem_type, threshold=0.50):
self.problem_type = handle_problem_types(problem_type)
if threshold < 0 or threshold > 1:
raise ValueError("threshold must be a float between 0 and 1, inclusive.")
self.threshold = threshold
[docs] def validate(self, X, y=None):
"""Check if there are any columns in the input that are too unique in the case of classification problems or not unique enough in the case of regression problems.
Args:
X (pd.DataFrame, np.ndarray): Features.
y (pd.Series, np.ndarray): Ignored. Defaults to None.
Returns:
dict: dict with a DataCheckWarning if there are any too unique or not
unique enough columns.
Examples:
>>> import pandas as pd
Because the problem type is regression, the column "regression_not_unique_enough" raises a warning
for having just one value.
>>> df = pd.DataFrame({
... "regression_unique_enough": [float(x) for x in range(100)],
... "regression_not_unique_enough": [float(1) for x in range(100)]
... })
...
>>> uniqueness_check = UniquenessDataCheck(problem_type="regression", threshold=0.8)
>>> assert uniqueness_check.validate(df) == [
... {
... "message": "Input columns 'regression_not_unique_enough' for regression problem type are not unique enough.",
... "data_check_name": "UniquenessDataCheck",
... "level": "warning",
... "code": "NOT_UNIQUE_ENOUGH",
... "details": {"columns": ["regression_not_unique_enough"], "uniqueness_score": {"regression_not_unique_enough": 0.0}, "rows": None},
... "action_options": [
... {
... "code": "DROP_COL",
... "parameters": {},
... "data_check_name": "UniquenessDataCheck",
... "metadata": {"columns": ["regression_not_unique_enough"], "rows": None}
... }
... ]
... }
... ]
For multiclass, the column "regression_unique_enough" has too many unique values and will raise
an appropriate warning.
>>> y = pd.Series([1, 1, 1, 2, 2, 3, 3, 3])
>>> uniqueness_check = UniquenessDataCheck(problem_type="multiclass", threshold=0.8)
>>> assert uniqueness_check.validate(df) == [
... {
... "message": "Input columns 'regression_unique_enough' for multiclass problem type are too unique.",
... "data_check_name": "UniquenessDataCheck",
... "level": "warning",
... "details": {
... "columns": ["regression_unique_enough"],
... "rows": None,
... "uniqueness_score": {"regression_unique_enough": 0.99}
... },
... "code": "TOO_UNIQUE",
... "action_options": [
... {
... "code": "DROP_COL",
... "data_check_name": "UniquenessDataCheck",
... "parameters": {},
... "metadata": {"columns": ["regression_unique_enough"], "rows": None}
... }
... ]
... }
... ]
...
>>> assert UniquenessDataCheck.uniqueness_score(y) == 0.65625
"""
messages = []
X = infer_feature_types(X)
res = X.apply(UniquenessDataCheck.uniqueness_score)
if is_regression(self.problem_type):
not_unique_enough_cols = list(res.index[res < self.threshold])
messages.append(
DataCheckWarning(
message=warning_not_unique_enough.format(
(", ").join(
["'{}'".format(str(col)) for col in not_unique_enough_cols],
),
self.problem_type,
),
data_check_name=self.name,
message_code=DataCheckMessageCode.NOT_UNIQUE_ENOUGH,
details={
"columns": not_unique_enough_cols,
"uniqueness_score": {
col: res.loc[col] for col in not_unique_enough_cols
},
},
action_options=[
DataCheckActionOption(
action_code=DataCheckActionCode.DROP_COL,
data_check_name=self.name,
metadata={"columns": not_unique_enough_cols},
),
],
).to_dict(),
)
elif is_multiclass(self.problem_type):
too_unique_cols = list(res.index[res > self.threshold])
messages.append(
DataCheckWarning(
message=warning_too_unique.format(
(", ").join(
["'{}'".format(str(col)) for col in too_unique_cols],
),
self.problem_type,
),
data_check_name=self.name,
message_code=DataCheckMessageCode.TOO_UNIQUE,
details={
"columns": too_unique_cols,
"uniqueness_score": {
col: res.loc[col] for col in too_unique_cols
},
},
action_options=[
DataCheckActionOption(
action_code=DataCheckActionCode.DROP_COL,
data_check_name=self.name,
metadata={"columns": too_unique_cols},
),
],
).to_dict(),
)
return messages
[docs] @staticmethod
def uniqueness_score(col, drop_na=True):
"""Calculate a uniqueness score for the provided field. NaN values are not considered as unique values in the calculation.
Based on the Herfindahl-Hirschman Index.
Args:
col (pd.Series): Feature values.
drop_na (bool): Whether to drop null values when computing the uniqueness score. Defaults to True.
Returns:
(float): Uniqueness score.
"""
norm_counts = (
col.value_counts(dropna=drop_na) / col.value_counts(dropna=drop_na).sum()
)
square_counts = norm_counts**2
score = 1 - square_counts.sum()
return score
