1+
2+ import logging
3+ import sys
4+ import json
5+
6+ def _config_str_to_dict (input_str ):
7+ # e.g.
8+ # input_str = "precision_recall_pair_threshold: 0.92,f_score_param: 0.5"
9+ # result_dict = {'precision_recall_pair_threshold': 0.92, 'f_score_param': 0.5}
10+
11+
12+ # Split the string into key-value pairs
13+ pairs = input_str .split (',' )
14+
15+ # Initialize an empty dictionary
16+ result_dict = {}
17+
18+ # Process each key-value pair
19+ for pair in pairs :
20+ # Split the pair by the colon
21+ key , value = pair .split (':' )
22+ # Remove any leading/trailing whitespace
23+ key = key .strip ()
24+ value = value .strip ()
25+ # Convert the value to float if possible
26+ try :
27+ value = float (value )
28+ except ValueError :
29+ pass # Keep the value as a string if it cannot be converted
30+ # Add the key-value pair to the dictionary
31+ result_dict [key ] = value
32+
33+ return result_dict
34+
35+
36+ def read_clusters (cluster_file ):
37+ """
38+ Reads the clusters from a file and returns a dictionary mapping node IDs to a set of cluster IDs.
39+ """
40+ node_to_clusters = {}
41+ with open (cluster_file , 'r' ) as f :
42+ for cluster_id , line in enumerate (f ):
43+ nodes = line .strip ().split ("\t " )
44+ for node in nodes :
45+ node = node .strip ()
46+ if node :
47+ if node not in node_to_clusters :
48+ node_to_clusters [node ] = set ()
49+ node_to_clusters [node ].add (cluster_id )
50+ return node_to_clusters
51+
52+ def read_ground_truth_pairs (ground_truth_file ):
53+ """
54+ Reads the ground truth pairs from a file and returns a list of tuples (node1, node2, weight).
55+ """
56+ pairs = []
57+ with open (ground_truth_file , 'r' ) as f :
58+ for line in f :
59+ parts = line .strip ().split ('\t ' )
60+ if len (parts ) >= 3 :
61+ node1 = parts [0 ].strip ()
62+ node2 = parts [1 ].strip ()
63+ try :
64+ weight = float (parts [2 ].strip ())
65+ except ValueError :
66+ print (f"Invalid weight '{ parts [2 ]} { line .strip ()} )
67+ continue
68+ pairs .append ((node1 , node2 , weight ))
69+ else :
70+ print (f"Ignoring invalid line: { line .strip ()} )
71+ return pairs
72+
73+ def compute_precision_recall (node_to_clusters , pairs , threshold ):
74+ """
75+ Computes precision and recall based on the clusters and ground truth pairs.
76+ Handles overlapping clusters where a node can belong to multiple clusters.
77+
78+ node_to_clusters: map from node id to a set of clusters
79+ pairs: list of (u,v,w) triplets
80+ """
81+ TP = 0 # True Positive
82+ TN = 0 # True Negative
83+ FP = 0 # False Positive
84+ FN = 0 # False Negative
85+
86+ for node1 , node2 , weight in pairs :
87+ # Determine if the pair is positive or negative
88+ is_positive = weight > threshold
89+
90+ # Determine if the nodes are in the same cluster (overlapping clusters)
91+ if node1 in node_to_clusters and node2 in node_to_clusters :
92+ clusters1 = node_to_clusters [node1 ]
93+ clusters2 = node_to_clusters [node2 ]
94+ in_same_cluster = bool (clusters1 & clusters2 ) # Check for non-empty intersection
95+ else :
96+ logging .warning ("skipping nodes %s, %s" , node1 , node2 )
97+ # Nodes not found in clusters; skip this pair
98+ continue
99+
100+ if is_positive :
101+ if in_same_cluster :
102+ TP += 1
103+ else :
104+ FN += 1
105+ else :
106+ if not in_same_cluster :
107+ TN += 1
108+ else :
109+ FP += 1
110+
111+ # Calculate precision and recall
112+ precision = TP / (TP + FP ) if (TP + FP ) > 0 else 0
113+ recall = TP / (TP + FN ) if (TP + FN ) > 0 else 0
114+
115+ return precision , recall , TP , FP , TN , FN
116+
117+
118+ def compute_precision_recall_pair (in_clustering , input_communities , out_statistics , stats_config , stats_dict ):
119+ """
120+ Compute pair precision and recall, and record result into stats_dict
121+ """
122+ stats_config = _config_str_to_dict (stats_config )
123+ precision_recall_pair_threshold = stats_config ["precision_recall_pair_threshold" ]
124+ f_score_param = stats_config .get ("f_score_param" , 1 )
125+ print ()
126+ print ("clustering file" , in_clustering )
127+ print ("community file" , input_communities )
128+ print ("stats file" , out_statistics )
129+ print ("parameters, " , precision_recall_pair_threshold , f_score_param )
130+
131+ # Read clusters and ground truth pairs
132+ clusters = read_clusters (in_clustering )
133+ pairs = read_ground_truth_pairs (input_communities )
134+
135+ # Compute precision and recall
136+ precision , recall , TP , FP , TN , FN = compute_precision_recall (clusters , pairs , precision_recall_pair_threshold )
137+
138+ f_score = 0
139+ if precision != 0 and recall != 0 :
140+ f_score = (1 + f_score_param * f_score_param ) * precision * recall / ((f_score_param * f_score_param * precision ) + recall )
141+
142+ stats_dict ["fScore_mean" ] = f_score
143+ stats_dict ["communityPrecision_mean" ] = precision
144+ stats_dict ["communityRecall_mean" ] = recall
145+ stats_dict ["PrecisionRecallPairThreshold" ] = precision_recall_pair_threshold
146+ stats_dict ["fScoreParam" ] = f_score_param
147+
148+ with open (out_statistics , 'w' , encoding = 'utf-8' ) as f :
149+ json .dump (stats_dict , f , indent = 4 )
0 commit comments