-
Notifications
You must be signed in to change notification settings - Fork 3
/
java-sim-bow2-opt.py
142 lines (122 loc) · 6.62 KB
/
java-sim-bow2-opt.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
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
# -*- coding: utf-8 -*-
"""
Bag-of-Words Similarity Detection for Java Code (Alternative implementation)
Martinez-Gil, J. (2024). Source Code Clone Detection Using Unsupervised Similarity Measures. arXiv preprint arXiv:2401.09885.
@author: Jorge Martinez-Gil
"""
import os
from sklearn.feature_extraction.text import CountVectorizer
from sklearn.metrics.pairwise import cosine_similarity
import javalang
def extract_semantic_tokens(code_snippet):
tokens = list(javalang.tokenizer.tokenize(code_snippet))
semantic_tokens = []
for token in tokens:
if isinstance(token, javalang.tokenizer.Identifier):
semantic_tokens.append(f"Identifier:{token.value}")
elif isinstance(token, javalang.tokenizer.Keyword):
semantic_tokens.append(f"Keyword:{token.value}")
elif isinstance(token, javalang.tokenizer.Operator):
semantic_tokens.append(f"Operator:{token.value}")
elif isinstance(token, javalang.tokenizer.Literal):
semantic_tokens.append(f"Literal:{token.value}")
else:
semantic_tokens.append(f"Other:{token.value}")
return ' '.join(semantic_tokens)
# Define the path to the IR-Plag-Dataset folder
dataset_path = os.path.join(os.getcwd(), "IR-Plag-Dataset")
# Define a list of similarity thresholds to iterate over
similarity_thresholds = [0.1, 0.3, 0.6]
# Initialize variables to keep track of the best result
best_threshold = 0
best_accuracy = 0
# Initialize counters
TP = 0
FP = 0
FN = 0
# Loop through each similarity threshold and calculate accuracy
for SIMILARITY_THRESHOLD in similarity_thresholds:
# Initialize the counters
total_cases = 0
over_threshold_cases_plagiarized = 0
over_threshold_cases_non_plagiarized = 0
cases_plag = 0
cases_non_plag = 0
# Loop through each subfolder in the dataset
for folder_name in os.listdir(dataset_path):
folder_path = os.path.join(dataset_path, folder_name)
if os.path.isdir(folder_path):
# Find the Java file in the original folder
original_path = os.path.join(folder_path, 'original')
java_files = [f for f in os.listdir(original_path) if f.endswith('.java')]
if len(java_files) == 1:
java_file = java_files[0]
with open(os.path.join(original_path, java_file), 'r') as f:
code1 = f.read()
# print(f"Found {java_file} in {original_path} for {folder_name}")
# Loop through each subfolder in the plagiarized and non-plagiarized folders
for subfolder_name in ['plagiarized', 'non-plagiarized']:
subfolder_path = os.path.join(folder_path, subfolder_name)
if os.path.isdir(subfolder_path):
# Loop through each Java file in the subfolder
for root, dirs, files in os.walk(subfolder_path):
for java_file in files:
if java_file.endswith('.java'):
with open(os.path.join(root, java_file), 'r') as f:
code2 = f.read()
# print(f"Found {java_file} in {root} for {folder_name}")
# Calculate the similarity ratio
semantic_tokens1 = extract_semantic_tokens(code1)
semantic_tokens2 = extract_semantic_tokens(code2)
vectorizer = CountVectorizer()
dtm = vectorizer.fit_transform([semantic_tokens1, semantic_tokens2])
similarity_matrix = cosine_similarity(dtm)
similarity_ratio = similarity_matrix[0, 1]
#print(f"{subfolder_name},{similarity_ratio:.2f}")
# Update the counters based on the similarity ratio
if subfolder_name == 'plagiarized':
cases_plag += 1
if similarity_ratio >= SIMILARITY_THRESHOLD:
over_threshold_cases_plagiarized += 1
elif subfolder_name == 'non-plagiarized':
cases_non_plag += 1
if similarity_ratio <= SIMILARITY_THRESHOLD:
over_threshold_cases_non_plagiarized += 1
total_cases += 1
# Update the counters based on the similarity ratio
if subfolder_name == 'plagiarized':
cases_plag += 1
if similarity_ratio >= SIMILARITY_THRESHOLD:
TP += 1 # True positive: plagiarized and identified as plagiarized
else:
FN += 1 # False negative: plagiarized but identified as non-plagiarized
elif subfolder_name == 'non-plagiarized':
cases_non_plag += 1
if similarity_ratio <= SIMILARITY_THRESHOLD:
over_threshold_cases_non_plagiarized += 1
else:
FP += 1 # False positive: non-plagiarized but identified as plagiarized
else:
print(f"Error: Found {len(java_files)} Java files in {original_path} for {folder_name}")
# Calculate accuracy for the current threshold
if total_cases > 0:
accuracy = (over_threshold_cases_non_plagiarized + over_threshold_cases_plagiarized) / total_cases
if accuracy > best_accuracy:
best_accuracy = accuracy
best_threshold = SIMILARITY_THRESHOLD
# Calculate precision and recall
if TP + FP > 0:
precision = TP / (TP + FP)
else:
precision = 0
if TP + FN > 0:
recall = TP / (TP + FN)
else:
recall = 0
# Calculate F-measure
if precision + recall > 0:
f_measure = 2 * (precision * recall) / (precision + recall)
else:
f_measure = 0
# Print the best threshold and accuracy
print(f"{os.path.basename(__file__)} - The best threshold is {best_threshold} with an accuracy of {best_accuracy:.2f}, Precision: {precision:.2f}, Recall: {recall:.2f}, F-measure: {f_measure:.2f}")