24161643270

Committed 08 Apr 2026 10:19PM UTC coverage: 74.52% (-0.2%) from 74.706%

Build # 24161643270

Build Type

Pull #340

github

Committed by

web-flow

Commit Message

Merge d52d63a47 into f7f8dd433

Pull Request Pull Request #340: feat: add feature importance classifier

Coverage Stats

119 of 182 new or added lines in 4 files covered. (65.38%)

6797 of 9121 relevant lines covered (74.52%)

0.75 hits per line

Source File
Press 'n' to go to next uncovered line, 'b' for previous

14.29

/src/capymoa/feature_selection/visualization.py

1	from __future__ import annotations	1✔
2
3	from typing import Sequence	1✔
4
5	import matplotlib.pyplot as plt	1✔
6	import pandas as pd	1✔
7
8
9	def _normalize_feature_names(	1✔
10	feature_names: Sequence[str] \| str \| None, n_features: int
11	) -> list[str]:
NEW 12	if feature_names is None:	×
NEW 13	return [f"f{i}" for i in range(n_features)]	×
NEW 14	if isinstance(feature_names, str):	×
NEW 15	names = [feature_names]	×
16	else:
NEW 17	names = [str(name) for name in feature_names]	×
18
NEW 19	if len(names) != n_features:	×
NEW 20	raise ValueError(	×
21	f"feature_names length ({len(names)}) does not match importances length ({n_features})."
22	)
NEW 23	return names	×
24
25
26	def plot_feature_importance(	1✔
27	importances: Sequence[float],
28	feature_names: Sequence[str] \| str \| None = None,
29	*,
30	top_k: int \| None = None,
31	ax=None,
32	title: str = "Feature importances",
33	):
34	"""Plot feature importances as a bar chart."""
NEW 35	values = list(importances)	×
NEW 36	names = _normalize_feature_names(feature_names, len(values))	×
37
NEW 38	series = pd.Series(values, index=names, name="importance").sort_values(	×
39	ascending=False
40	)
NEW 41	if top_k is not None:	×
NEW 42	if top_k <= 0:	×
NEW 43	raise ValueError("top_k must be greater than zero.")	×
NEW 44	series = series.head(top_k)	×
45
NEW 46	if ax is None:	×
NEW 47	_, ax = plt.subplots(figsize=(10, 4))	×
48
NEW 49	series.plot(kind="bar", ax=ax)	×
NEW 50	ax.set_title(title)	×
NEW 51	ax.set_xlabel("Feature")	×
NEW 52	ax.set_ylabel("Importance")	×
NEW 53	ax.figure.tight_layout()	×
NEW 54	return ax	×
55
56
57	def plot_windowed_feature_importance(	1✔
58	windowed_importances: list[dict],
59	feature_names: Sequence[str] \| str \| None = None,
60	*,
61	top_k: int \| None = None,
62	ax=None,
63	title: str = "Windowed feature importances",
64	):
65	"""Plot windowed feature importances over time."""
NEW 66	if len(windowed_importances) == 0:	×
NEW 67	raise ValueError("windowed_importances is empty.")	×
NEW 68	if top_k is not None and top_k <= 0:	×
NEW 69	raise ValueError("top_k must be greater than zero.")	×
70
NEW 71	n_features = len(windowed_importances[0]["importances"])	×
NEW 72	names = _normalize_feature_names(feature_names, n_features)	×
73
NEW 74	window_df = pd.DataFrame(	×
75	[entry["importances"] for entry in windowed_importances],
76	columns=names,
77	)
NEW 78	window_df["instances_seen"] = [	×
79	entry["instances_seen"] for entry in windowed_importances
80	]
81
NEW 82	if top_k is None:	×
NEW 83	top_features = names	×
84	else:
NEW 85	top_features = (	×
86	window_df[names]
87	.mean()
88	.sort_values(ascending=False)
89	.head(top_k)
90	.index.tolist()
91	)
92
NEW 93	if ax is None:	×
NEW 94	_, ax = plt.subplots(figsize=(10, 4))	×
95
NEW 96	window_df.plot(x="instances_seen", y=top_features, ax=ax)	×
NEW 97	ax.set_title(title)	×
NEW 98	ax.set_xlabel("Instances seen")	×
NEW 99	ax.set_ylabel("Importance")	×
NEW 100	ax.figure.tight_layout()	×
NEW 101	return ax	×

adaptive-machine-learning / CapyMOA / 24161643270

Source File Press 'n' to go to next uncovered line, 'b' for previous

Source File
Press 'n' to go to next uncovered line, 'b' for previous