-
Notifications
You must be signed in to change notification settings - Fork 0
/
clustering_metrics.py
71 lines (60 loc) · 2.44 KB
/
clustering_metrics.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
import seaborn as sns
import sklearn.metrics
def _draw_heatmap(confusion_matrix):
"""
Draws a confusion matrix as a heatmap.
Parameters:
-----------
confusion_matrix : confusion_matrix object
The output of sklearn.metrics.confusion_matrix().
"""
heatmap = sns.heatmap(confusion_matrix, cmap='magma')
heatmap.set(xlabel='clusters', ylabel='true labels')
def evaluate_clustering(data, true_labels, cluster_assignments, heatmap=True):
"""
Evaluates the clustering performance given the cluster assignments.
Also, calls _draw_heatmap() on the confusion matrix of the cluster assignments vs the true labels.
Parameters:
-----------
data : array_like
Data to evaluate the model.
true_labels : list
True labels of the data.
cluster_assignments : array_like
Cluster assignments of the data.
Returns:
--------
return : dictionary
A dictionary containing some performance metrics.
"""
confusion_matrix = sklearn.metrics.confusion_matrix(true_labels, cluster_assignments)
matched_labels = confusion_matrix.argmax(0)[cluster_assignments]
if heatmap:
_draw_heatmap(confusion_matrix)
return {"Acc" : sklearn.metrics.accuracy_score(true_labels, matched_labels),
"ARI" : sklearn.metrics.adjusted_rand_score(true_labels, matched_labels),
"AMI" : sklearn.metrics.adjusted_mutual_info_score(true_labels, matched_labels),
"Sil" : sklearn.metrics.silhouette_score(data, cluster_assignments),
}
def evaluate_model(data, true_labels, clustering_method, encode_method, heatmap=True):
"""
Evaluates the model clustering performance:
Computes the cluster assignment and embeddings and calls evaluate_clustering().
Parameters:
-----------
data : array_like
Data to evaluate the model.
true_labels : list
True labels of the data.
clustering_method : method
Model method (or function) which returns the cluster assignments.
encode_method : method
Model method (or function) which returns the embeddings on the latent space.
Returns:
--------
return : dictionary
A dictionary containing some performance metrics.
"""
embeddings = encode_method(data)
cluster_assignments = clustering_method(data)
return evaluate_clustering(embeddings, true_labels, cluster_assignments, heatmap=heatmap)