def test_automatically_find_variables_and_return_as_numeric(df_normal_dist):
# test case 1: automatically select variables, return_object=False
transformer = EqualWidthDiscretiser(bins=10,
variables=None,
return_object=False)
X = transformer.fit_transform(df_normal_dist)
# fit parameters
_, bins = pd.cut(x=df_normal_dist["var"],
bins=10,
retbins=True,
duplicates="drop")
bins[0] = float("-inf")
bins[len(bins) - 1] = float("inf")
# transform output
X_t = [x for x in range(0, 10)]
val_counts = [18, 17, 16, 13, 11, 7, 7, 5, 5, 1]
# init params
assert transformer.bins == 10
assert transformer.variables == ["var"]
assert transformer.return_object is False
# fit params
assert transformer.input_shape_ == (100, 1)
# transform params
assert (transformer.binner_dict_["var"] == bins).all()
assert len([x for x in X["var"].unique() if x not in X_t]) == 0
# in equal width discretisation, intervals get different number of values
assert len([x for x in X["var"].value_counts()
if x not in val_counts]) == 0
def test_automatically_find_variables_and_return_as_object(df_normal_dist):
transformer = EqualWidthDiscretiser(bins=10,
variables=None,
return_object=True)
X = transformer.fit_transform(df_normal_dist)
assert X["var"].dtypes == "O"
def fit(self, X: pd.DataFrame, y: pd.Series = None):
"""
Find features with high PSI values.
Parameters
----------
X : pandas dataframe of shape = [n_samples, n_features]
The training dataset.
y : pandas series. Default = None
y is not needed in this transformer. You can pass y or None.
"""
# check input dataframe
X = check_X(X)
# If required exclude variables that are not in the input dataframe
self._confirm_variables(X)
# find numerical variables or check those entered are present in the dataframe
self.variables_ = _find_or_check_numerical_variables(
X, self.variables_)
# Remove the split_col from the variables list. It might be added if the
# variables are not defined at initialization.
if self.split_col in self.variables_:
self.variables_.remove(self.split_col)
if self.missing_values == "raise":
# check if dataset contains na or inf
_check_contains_na(X, self.variables_)
_check_contains_inf(X, self.variables_)
# Split the dataframe into basis and test.
basis_df, test_df = self._split_dataframe(X)
# Check the shape of the returned dataframes for PSI calculations.
# The number of observations must be at least equal to the
# number of bins.
if min(basis_df.shape[0], test_df.shape[0]) < self.bins:
raise ValueError(
"The number of rows in the basis and test datasets that will be used "
f"in the PSI calculations must be at least larger than {self.bins}. "
"After slitting the original dataset based on the given cut_off or"
f"split_frac we have {basis_df.shape[0]} samples in the basis set, "
f"and {test_df.shape[0]} samples in the test set. "
"Please adjust the value of the cut_off or split_frac.")
# Switch basis and test dataframes if required.
if self.switch:
test_df, basis_df = basis_df, test_df
# set up the discretizer
if self.strategy == "equal_width":
bucketer = EqualWidthDiscretiser(bins=self.bins)
else:
bucketer = EqualFrequencyDiscretiser(q=self.bins)
# Compute the PSI by looping over the features
self.psi_values_ = {}
self.features_to_drop_ = []
for feature in self.variables_:
# Discretize the features.
basis_discrete = bucketer.fit_transform(basis_df[[feature
]].dropna())
test_discrete = bucketer.transform(test_df[[feature]].dropna())
# Determine percentage of observations per bin
basis_distrib, test_distrib = self._observation_frequency_per_bin(
basis_discrete, test_discrete)
# Calculate the PSI value
self.psi_values_[feature] = np.sum(
(test_distrib - basis_distrib) *
np.log(test_distrib / basis_distrib))
# Assess if feature should be dropped
if self.psi_values_[feature] > self.threshold:
self.features_to_drop_.append(feature)
# save input features
self._get_feature_names_in(X)
return self