def test_fewer_bins_than_n_discretize_bins():
df = pd.read_csv("credit.csv")
for n in range(2, 20, 5):
bin_transformer_ = BinTransformer(n_discretize_bins=n)
bin_transformer_.fit(df)
for feat, bin_ranges in bin_transformer_.bins_.items():
assert len(bin_ranges) <= n
def test_correct_min_max_bins():
df = pd.read_csv("credit.csv")
bin_transformer_ = BinTransformer()
bin_transformer_.fit(df)
for feat, bins in bin_transformer_.bins_.items():
assert bins[0][0] == df[feat].min()
assert bins[-1][1] == df[feat].max()
def test_each_bin_in_order():
df = pd.read_csv("credit.csv")
bin_transformer_ = BinTransformer()
bin_transformer_.fit(df)
for feat, bins in bin_transformer_.bins_.items():
for i, bin in enumerate(bins):
assert (bin[0] <= bin[1])
def test_bin_ranges_are_flush():
df = pd.read_csv("credit.csv")
bin_transformer_ = BinTransformer()
bin_transformer_.fit(df)
for feat, bin_ranges in bin_transformer_.bins_.items():
prev_ceil = None
for floor, ceil in bin_ranges:
assert prev_ceil is None or floor == prev_ceil
prev_ceil = ceil
def preprocess_training_data(preprocess_params):
# Get params
trainset = preprocess_params["trainset"]
y = preprocess_params["y"]
class_feat = preprocess_params["class_feat"]
pos_class = preprocess_params["pos_class"]
user_requested_feature_names = preprocess_params["feature_names"]
n_discretize_bins = preprocess_params["n_discretize_bins"]
verbosity = preprocess_params["verbosity"]
# Error check
_check_valid_input_data(
trainset,
y,
class_feat,
user_requested_feature_names=user_requested_feature_names,
)
# Determine class_feat
class_feat = _get_class_feat_name(class_feat, y)
# Build new DataFrame containing both X and y.
df = _convert_to_training_df(
trainset,
y,
class_feat,
user_requested_feature_names=user_requested_feature_names,
)
# Define pos_class
pos_class = _get_pos_class(df, class_feat, pos_class)
# Infer correct datatypes
df = df.infer_objects()
# Bin, if necessary
bin_transformer_ = BinTransformer(n_discretize_bins=n_discretize_bins,
verbosity=verbosity)
df = bin_transformer_.fit_transform(df, ignore_feats=[class_feat])
# Done
return df, class_feat, pos_class, bin_transformer_
def test_no_bins():
old_df = pd.read_csv("credit.csv")
df = old_df.copy()
bin_transformer_ = BinTransformer(n_discretize_bins=0)
bin_transformer_.fit(df)
bin_transformer_.transform(df)
assert df.equals(old_df)
def _upgrade_bin_transformer_ifdepr(obj):
old_bin_transformer_ = getattr(obj, "bin_transformer_")
if type(old_bin_transformer_) == dict:
new_bin_transformer_ = BinTransformer()
new_bin_transformer_.bins_ = old_bin_transformer_
setattr(obj, "bin_transformer_", new_bin_transformer_)