def plot_lift_curve(
y_true: DataType,
y_proba: DataType,
title: str = None,
ax: Axes = None,
labels: List[str] = None,
threshold: float = 0.5,
) -> Axes:
"""
Plot a lift chart from results. Also calculates lift score based on a .5 threshold
Parameters
----------
y_true: DataType
True labels
y_proba: DataType
Model's predicted probability
title: str
Plot title
ax: Axes
Pass your own ax
labels: List of str
Labels to use per class
threshold: float
Threshold to use when determining lift score
Returns
-------
matplotlib.Axes
"""
if ax is None:
fig, ax = plt.subplots()
title = "Lift Curve" if title is None else title
classes = np.unique(y_true)
binarized_labels = label_binarize(y_true, classes=classes)
if labels and len(labels) != len(classes):
raise VizError("Number of labels must match number of classes: "
f"got {len(labels)} labels and {len(classes)} classes")
if binarized_labels.shape[1] == 1:
# Binary classification case
percents, gains = _cum_gain_curve(binarized_labels, y_proba[:, 1])
score = lift_score(binarized_labels.ravel(), y_proba[:, 1] > threshold)
ax.plot(percents, gains / percents, label=f"$Lift = {score:.2f}$")
else:
# Multi-class case
for class_ in classes:
percents, gains = _cum_gain_curve(binarized_labels[:, class_],
y_proba[:, class_])
score = lift_score(binarized_labels[:, class_],
y_proba[:, class_] > threshold)
ax.plot(
percents,
gains / percents,
label=f"Class {labels[class_] if labels else class_} "
f"$Lift = {score:.2f}$ ",
)
ax.axhline(y=1, color="grey", linestyle="--", label="Baseline")
ax.set_title(title)
ax.set_ylabel("Lift")
ax.set_xlabel("% of Data")
formatter = PercentFormatter(xmax=1)
ax.xaxis.set_major_formatter(formatter)
ax.legend()
return ax
def test_lift_score_returns_correctly(self, y_targ, y_pred):
result = lift_score(y_targ, y_pred)
assert 2 == result
def test_lift_score_fails_if_passed_non_ndarray():
with pytest.raises(MetricError):
lift_score([1, 2, 3], [4, 5, 6])
def test_lift_score_fails_if_passed_non_ndarray_or_series(
self, y_targ, y_pred):
with pytest.raises(MetricError):
# noinspection PyTypeChecker
lift_score(y_targ, y_pred)
def test_lift_score_returns_correctly():
y_targ = np.array([1, 1, 1, 0, 0, 2, 0, 3, 4])
y_pred = np.array([1, 0, 1, 0, 0, 2, 1, 3, 0])
result = lift_score(y_targ, y_pred)
assert 2 == result