Add functions for finding ancestors and calculating hierarchical precision and recall
Browse files
ham.py
CHANGED
@@ -1,4 +1,7 @@
|
|
1 |
-
|
|
|
|
|
|
|
2 |
"""
|
3 |
Recursively finds ancestors of a given class (e.g., an ISCO-08 code) in a hierarchical JSON structure.
|
4 |
|
@@ -19,6 +22,50 @@ def find_ancestors(tree, code):
|
|
19 |
return ancestors
|
20 |
|
21 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
22 |
def calculate_hierarchical_measures(true_labels, predicted_labels, tree):
|
23 |
"""
|
24 |
Calculates hierarchical precision, recall, and F-measure in a hierarchical structure.
|
|
|
1 |
+
from typing import List, Set, Dict, Tuple
|
2 |
+
|
3 |
+
|
4 |
+
def find_ancestors1(tree, code):
|
5 |
"""
|
6 |
Recursively finds ancestors of a given class (e.g., an ISCO-08 code) in a hierarchical JSON structure.
|
7 |
|
|
|
22 |
return ancestors
|
23 |
|
24 |
|
25 |
+
def find_ancestors(node, hierarchy):
|
26 |
+
ancestors = set()
|
27 |
+
nodes_to_visit = [node]
|
28 |
+
while nodes_to_visit:
|
29 |
+
current_node = nodes_to_visit.pop()
|
30 |
+
if current_node in hierarchy:
|
31 |
+
parents = hierarchy[current_node]
|
32 |
+
ancestors.update(parents)
|
33 |
+
nodes_to_visit.extend(parents)
|
34 |
+
return ancestors
|
35 |
+
|
36 |
+
|
37 |
+
def extend_with_ancestors(classes, hierarchy):
|
38 |
+
extended_classes = set(classes)
|
39 |
+
for cls in classes:
|
40 |
+
ancestors = find_ancestors(cls, hierarchy)
|
41 |
+
extended_classes.update(ancestors)
|
42 |
+
return extended_classes
|
43 |
+
|
44 |
+
|
45 |
+
def calculate_hierarchical_precision_recall(
|
46 |
+
real_codes: List[str], predicted_codes: List[str], hierarchy: Dict[str, Set[str]]
|
47 |
+
) -> Tuple[float, float]:
|
48 |
+
# Extend the sets of real and predicted codes with their ancestors
|
49 |
+
extended_real = set()
|
50 |
+
for code in real_codes:
|
51 |
+
extended_real.add(code)
|
52 |
+
extended_real.update(hierarchy.get(code, set()))
|
53 |
+
|
54 |
+
extended_predicted = set()
|
55 |
+
for code in predicted_codes:
|
56 |
+
extended_predicted.add(code)
|
57 |
+
extended_predicted.update(hierarchy.get(code, set()))
|
58 |
+
|
59 |
+
# Calculate the intersection
|
60 |
+
correct_predictions = extended_real.intersection(extended_predicted)
|
61 |
+
|
62 |
+
# Calculate hierarchical precision and recall
|
63 |
+
hP = len(correct_predictions) / len(extended_predicted) if extended_predicted else 0
|
64 |
+
hR = len(correct_predictions) / len(extended_real) if extended_real else 0
|
65 |
+
|
66 |
+
return hP, hR
|
67 |
+
|
68 |
+
|
69 |
def calculate_hierarchical_measures(true_labels, predicted_labels, tree):
|
70 |
"""
|
71 |
Calculates hierarchical precision, recall, and F-measure in a hierarchical structure.
|