def test_di_multiple_features_multiple_rules(self):
"""Tests that correct distance is computed for 2 numeric features"""
df = pd.DataFrame({"A": [1, 1, 4, 1.5, 0.5, 0.75], "B": [3, 2, 1, .5, 3, 2],
"C": ["a", "b", "c", "d", "e", "f"],
"Class": ["apple", "apple", "banana", "banana", "banana", "banana"]})
class_col_name = "Class"
rules = [pd.Series({"A": (1, 2), "B": (1, 2), "Class": "banana"}),
pd.Series({"A": (1, 2), "B": (1, 2), "Class": "banana"})]
dists = []
min_max = pd.DataFrame({"A": {"min": 1, "max": 5}, "B": {"min": 1, "max": 11}})
# Only keep rows with the same class label as the rule
df = df.loc[df[class_col_name] == "banana"]
correct = [pd.Series({2: 0.5*0.5, 3: 0.0, 4: 0.125*0.125, 5: 1/16*1/16}),
pd.Series({2: 0, 3: 0.05 * 0.05, 4: 0.1*0.1, 5: 0.0}),
pd.Series({2: 0.5 * 0.5, 3: 0.0, 4: 0.125 * 0.125, 5: 1 / 16 * 1 / 16}),
pd.Series({2: 0, 3: 0.05 * 0.05, 4: 0.1 * 0.1, 5: 0.0})]
for rule in rules:
for i, col_name in enumerate(df):
if col_name == class_col_name:
continue
col = df[col_name]
if is_numeric_dtype(col):
dist = di(col, rule, min_max)
dists.append(dist)
self.assertTrue(dists[i].equals(correct[i]))
def test_di_nan_rule(self):
"""Tests that correct distance is computed if NaNs occur in a rule"""
df = pd.DataFrame({"A": ["high", np.nan, "high", "low", "low", "high"], "B": [3, 2, 1, np.nan, 1, 2],
"Class": ["apple", "apple", "banana", "banana", "banana", "banana"]})
class_col_name = "Class"
rule = pd.Series({"A": "high", "B": (np.NaN, 2), "Class": "banana"})
min_max = pd.DataFrame({"B": {"min": 1, "max": 2}})
correct = pd.Series([1.0, 1.0, 1.0, 1.0, 1.0, 1.0], name="A")
for i, col_name in enumerate(df):
if col_name == class_col_name:
continue
col = df[col_name]
if is_numeric_dtype(col):
dist = di(col, rule, min_max)
self.assertTrue(dist.equals(correct))
def test_di_single_feature(self):
"""Tests that correct distance is computed for 1 numeric feature"""
df = pd.DataFrame({"A": ["high", np.nan, "high", "low", "low", "high"], "B": [3, 2, 1, .5, 1, 2],
"Class": ["apple", "apple", "banana", "banana", "banana", "banana"]})
class_col_name = "Class"
rule = pd.Series({"A": "high", "B": (1, 2), "Class": "banana"})
dist = None
min_max = pd.DataFrame({"B": {"min": 1, "max": 5}})
correct = pd.Series({0: 0.25*0.25, 1: 0, 2: 0, 3: 0.125*0.125, 4: 0, 5: 0.0})
for i, col_name in enumerate(df):
if col_name == class_col_name:
continue
col = df[col_name]
if is_numeric_dtype(col):
dist = di(col, rule, min_max)
self.assertTrue(dist.equals(correct))
def test_di_nan_row(self):
"""Tests that correct distance is computed if NaNs occur in a row of a column"""
df = pd.DataFrame({"A": ["high", np.nan, "high", "low", "low", "high"], "B": [3, 2, 1, np.nan, 0.5, 2],
"C": [np.nan, np.nan, np.nan, np.nan, np.nan, np.nan],
"Class": ["apple", "apple", "banana", "banana", "banana", "banana"]})
class_col_name = "Class"
rule = pd.Series({"A": "high", "B": (1, 2), "C":(1, np.NaN), "Class": "banana"})
min_max = pd.DataFrame({"B": {"min": 1, "max": 5}})
correct = [pd.Series([1/4*1/4, 0.0, 0.0, 1.0, 1/8*1/8, 0.0], name="A"),
pd.Series([1.0, 1.0, 1.0, 1.0, 1.0, 1.0], name="A")]
j = 0
for i, col_name in enumerate(df):
if col_name == class_col_name:
continue
col = df[col_name]
if is_numeric_dtype(col):
dist = di(col, rule, min_max)
self.assertTrue(dist.equals(correct[j]))
j += 1