/
recommendation.py
153 lines (108 loc) · 4.87 KB
/
recommendation.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
143
144
145
146
147
148
149
150
151
152
import pandas as pd
import numpy as np
import sklearn.metrics.pairwise as pairwise
from sklearn.feature_extraction.text import TfidfVectorizer,CountVectorizer
from lib.nlp import processed, lower, split
from lib.evaluate import evaluate_file
import warnings
warnings.simplefilter(action='ignore', category=Warning)
stemming_text=True
stemming_concept=False
binary_vector=False
is_tfidf=True
if is_tfidf:
binary_vector=False
def get_match(vec_file, cfile):
#Load Concepts
clist = pd.read_csv(cfile)
sec_or_q_to_concepts = clist.groupby('item_id')['concept'].apply(set).to_dict()
# Load Book Sections and text
df_book = pd.read_csv('data/section2text.csv', encoding='utf8')
book_ids = df_book['section'].tolist()
book_texts = df_book['text'].apply(lower).tolist()
# Quiz Text
df_quiz = pd.read_csv('data/quiz_text_section.csv').fillna('')
quiz_ids = df_quiz['quizid'].tolist()
quiz_texts = df_quiz['text'].apply(lower).tolist()
book_texts = [processed(i, stemming=stemming_text) for i in book_texts]
quiz_texts = [processed(i, stemming=stemming_text) for i in quiz_texts]
concept_all_list = list(set([j for i in sec_or_q_to_concepts for j in sec_or_q_to_concepts[i]]))
# Text Representation BoW approach
if is_tfidf:
vectorizer = TfidfVectorizer()
else:
vectorizer = CountVectorizer()
all_texts = book_texts + quiz_texts
vectorizer.fit_transform(all_texts)
Q_vector_content = vectorizer.transform(quiz_texts).toarray()
B_vector_content = vectorizer.transform(book_texts).toarray()
# Concept Representation Binary Vector Approach
B_vector_concept = [
[1 if i in sec_or_q_to_concepts[s] else 0 for i in concept_all_list] for s in book_ids
]
Q_vector_concept = [
[1 if i in sec_or_q_to_concepts[s] else 0 for i in concept_all_list] for s in quiz_ids]
dict_Q_vector_content = dict(zip(quiz_ids, Q_vector_content))
dict_Q_vector_concept = dict(zip(quiz_ids, Q_vector_concept))
df = pd.read_csv(vec_file)
features = df.columns[6:].tolist()
features = [processed(i,stemming=stemming_concept) for i in features]
df_q_0 = df[(df['quiz_id'].str.startswith('q')) & (df['outcome'].isin(('0', 0)))]
matrix_content = []
matrix_concept = []
matrix_knowledge = []
r = []
for row in df_q_0.values:
interid, type_, quiz_id, concepts, student_id, outcome, quiz_sec = row[:7]
content_vector = dict_Q_vector_content[quiz_id]
concept_vector = dict_Q_vector_concept[quiz_id]
concept_vector_knowledge = np.array(concept_vector, dtype=np.float64)
knowledge_v = row[7:]
concepts = sec_or_q_to_concepts[quiz_id]
for k in concepts:
if k in features:
i = features.index(k)
else:
i=-1
# print(k,"not in list")
if i >= 0:
v = knowledge_v[i]
if 0. <= v <= 1.0:
index = concept_all_list.index(k)
concept_vector_knowledge[index] = 1-v
matrix_content.append(content_vector)
matrix_concept.append(concept_vector)
matrix_knowledge.append(concept_vector_knowledge)
sim_matrix_content = pairwise.cosine_similarity(matrix_content, B_vector_content)
sim_matrix_concept = pairwise.cosine_similarity(matrix_concept, B_vector_concept)
sim_matrix_knowledge = pairwise.cosine_similarity(matrix_knowledge, B_vector_concept)
# find matches
def get_match_top_n(sim, n=5):
best = []
for row in sim:
tmp = []
for idx in reversed(row.argsort()[-n:]):
if row[idx] > 0.0:
tmp.append(book_ids[idx])
best.append(tmp)
return best
df_match = pd.DataFrame()
df_match['interaction_id'] = df_q_0['interaction_id']
df_match['content_match'] = get_match_top_n(sim_matrix_content)
df_match['concept_match'] = get_match_top_n(sim_matrix_concept)
df_match['knowledge_match'] = get_match_top_n(sim_matrix_knowledge)
alpha=0.6
sim = alpha * sim_matrix_knowledge + (1-alpha) * sim_matrix_content
df_match['sim_com_knowledge_norm_concat_{:.2f}_match'.format(alpha)] = get_match_top_n(sim)
sim = alpha * sim_matrix_knowledge + (1-alpha) * sim_matrix_concept
df_match['sim_com_knowledge_norm_concept_{:.2f}_match'.format(alpha)] = get_match_top_n(sim)
df = df[['interaction_id', 'quiz_id', 'outcome', 'student_id', 'quiz_section']]
df = df.merge(df_match, on='interaction_id', how='left')
return df
if __name__ == '__main__':
vec_file= 'data/vec/pfa_outcome_concept_list_expert.csv'
concept_file='data/concept_files/list_filter_stemmed_TopicRank_concepts.txt'
df = get_match(vec_file,concept_file)
df.to_csv("match.csv", index=False)
print('matching file-created',"data/match.csv")
evaluate_file('match.csv')