-
Notifications
You must be signed in to change notification settings - Fork 0
/
feature_attribution.py
77 lines (58 loc) · 3.42 KB
/
feature_attribution.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
"""Contains all functions related to the feature attribution"""
def calculate_very_basic_feature_attribution(anomaly: int, anomaly_data: dict) -> list[float]:
"""Calculates a feature attribution based on the specified anomaly and the output of the anomaly detection.
Args:
anomaly: The ID of the anomaly.
anomaly_data: The output of the anomaly detection.
Returns:
A percentage for each feature that determines its influence on the detected anomaly.
"""
anomaly_index = anomaly_data["anomalies"][anomaly]["index"]
return [anomaly_data["deep-error"][i][anomaly_index] / anomaly_data["error"][anomaly_index] * 100 for i in range(len(anomaly_data["sensors"]))]
def calculate_basic_feature_attribution(anomaly: int, anomaly_data: dict) -> list[float]:
"""Calculates a feature attribution based on the specified anomaly and the output of the anomaly detection.
Uses the middle of the anomaly area for the feature attribution
Args:
anomaly: The ID of the anomaly.
anomaly_data: The output of the anomaly detection.
Returns:
A percentage for each feature that determines its influence on the detected anomaly.
"""
anomaly_index = anomaly_data["anomalies"][anomaly]["index"] + anomaly_data["anomalies"][anomaly]["length"] // 2
return [anomaly_data["deep-error"][i][anomaly_index] / anomaly_data["error"][anomaly_index] * 100 for i in range(len(anomaly_data["sensors"]))]
def calculate_averaged_feature_attribution(anomaly: int, anomaly_data: dict) -> list[float]:
"""Calculates results feature attribution based on the specified anomaly and the output of the anomaly detection.
Uses the averaged results of the anomaly area for the feature attribution
Args:
anomaly: The ID of the anomaly.
anomaly_data: The output of the anomaly detection.
Returns:
A percentage for each feature that determines its influence on the detected anomaly.
"""
anomaly_index = anomaly_data["anomalies"][anomaly]["index"]
anomaly_length = anomaly_data["anomalies"][anomaly]["length"]
results = []
for i in range(len(anomaly_data["sensors"])):
sensor_percentages = []
for j in range(anomaly_index, anomaly_index + anomaly_length):
sensor_percentages.append(anomaly_data["deep-error"][i][j])
results.append(sum(sensor_percentages) / len(sensor_percentages))
return [(e / sum(results)) * 100 for e in results]
def calculate_median_feature_attribution(anomaly: int, anomaly_data: dict) -> list[float]:
"""Calculates a feature attribution based on the specified anomaly and the output of the anomaly detection.
Uses the median values of the anomaly area for the feature attribution
Args:
anomaly: The ID of the anomaly.
anomaly_data: The output of the anomaly detection.
Returns:
A percentage for each feature that determines its influence on the detected anomaly.
"""
anomaly_index = anomaly_data["anomalies"][anomaly]["index"]
anomaly_length = anomaly_data["anomalies"][anomaly]["length"]
results = []
for i in range(len(anomaly_data["sensors"])):
sensor_percentages = []
for j in range(anomaly_index, anomaly_index + anomaly_length):
sensor_percentages.append(anomaly_data["deep-error"][i][j])
results.append(sorted(sensor_percentages)[len(sensor_percentages) // 2])
return [(e / sum(results)) * 100 for e in results]