-
Notifications
You must be signed in to change notification settings - Fork 0
/
classifier.py
212 lines (147 loc) · 7.16 KB
/
classifier.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
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
# -------------------------------------------------------
# Assignment 2
# Written by Joshua Parial-Bolusan (40063663) Jeffrey Lam(40090989)
# For COMP 472 Section AA – Summer 2021
# --------------------------------------------------------
from typing import Dict
from service import imdb_service
from service.imdb_service import *
from math import log10
import re
import numpy as np
class Classifier:
def __init__(self, reviews_df):
self.reviews_df = reviews_df
def build_vocabulary(self, smooth=1):
reviews_df = self.reviews_df
positive_df = reviews_df[reviews_df["rating"] == "positive"]
negative_df = reviews_df[reviews_df["rating"] == "negative"]
pos_off = int(len(positive_df)/2)
neg_offset = int(len(negative_df)/2)
train_df = (positive_df.iloc[:pos_off]).append(negative_df.iloc[:neg_offset])
train_pos = train_df[train_df["rating"] == "positive"]
train_neg = train_df[train_df["rating"] == "negative"]
test_df = (positive_df.iloc[pos_off:]).append(negative_df.iloc[neg_offset:])
frequencies = {}
removed_words = []
stop_word_file = open("stopword.txt", "r")
stop_words = [word.strip() for word in stop_word_file.readlines()]
stop_word_file.close()
model = {
"word": [],
"positive": [],
"positive_prob": [],
"negative": [],
"negative_prob": []
}
#Count Frequencies
for index, row in train_df.iterrows():
for word in row["review"].split(" "):
if word in stop_words:
removed_words.append(word)
if word not in stop_words:
frequencies.setdefault(word, {"positive": 0, "negative": 0})
frequencies[word][row["rating"]] += 1
#write removed workds to remove.txt
self.remove_to_file(removed_words, "remove.txt")
words_in_pos = 0
words_in_neg = 0
vocab_size = len(frequencies.keys())
for key in frequencies.keys():
words_in_pos += frequencies[key]["positive"]
words_in_neg += frequencies[key]["negative"]
for word in frequencies.keys():
pos_freq = frequencies[word]["positive"]
neg_freq = frequencies[word]["negative"]
model["word"].append(word)
model["positive"].append(pos_freq)
model["positive_prob"].append((pos_freq + smooth) / (vocab_size + words_in_pos*smooth))
model["negative"].append(neg_freq)
model["negative_prob"].append((neg_freq + smooth) / (vocab_size + words_in_neg*smooth))
return pd.DataFrame(model), len(train_pos), len(train_neg), test_df
def modify_smooth(self, train_model: pd.DataFrame, smooth=1):
vocab_size = train_model.shape[0]
pos_count = train_model["positive"].sum()
neg_count = train_model["negative"].sum()
train_model["positive_prob"] = train_model.apply(lambda row: ((row.positive+ smooth) / (vocab_size + pos_count*smooth)), axis=1)
train_model["negative_prob"] = train_model.apply(lambda row: ((row.negative + smooth) / (vocab_size + neg_count*smooth)), axis=1)
return train_model
def modify_length(self, train_model: pd.DataFrame, length):
indexes_to_drop=[]
if (length <= 4):
#print(train_model.head(50))
for index in train_model.index:
word = train_model.iloc[index]["word"]
if (len(word) <= length):
#print(train_model.loc[index])
indexes_to_drop.append(index)
#train_model = train_model.drop(index, axis=0)
#train_model = train_model.drop(index= word)
elif (length >= 9):
for index, row in train_model.iterrows():
#print(train_model.loc[[index]])
if (len(row["word"]) >= length):
indexes_to_drop.append(index)
#train_model = train_model.drop(index, axis=0)
#train_model = train_model.drop(index =word)
train_model = train_model.drop(train_model.index[indexes_to_drop])
train_model.reset_index(drop=True, inplace=True)
return train_model
def remove_to_file(self, words, file: str):
remove_file = open(file, "w", encoding="utf-8")
for word in words:
remove_file.write(f"{word}\n")
remove_file.close()
def model_to_file(self, model: pd.DataFrame, file: str):
model_file = open(file, "w", encoding="utf-8")
for index, row in model.iterrows():
word = row["word"]
pos_f = row["positive"]
pos_prob = row["positive_prob"]
neg_f = row["negative"]
neg_prob = row["negative_prob"]
model_file.write(f"No.{index} {word}\n{pos_f}, {pos_prob}, {neg_f}, {neg_prob}\n")
model_file.close()
def results_to_file(self, results: pd.DataFrame, file):
results_file = open(file, "w", encoding="utf-8")
for index, row in results.iterrows():
title = row["title"]
pos_result = row["positive"]
neg_result = row["negative"]
result = row["result"]
actual_result = row["actual_result"]
outcome = "right" if row["prediction"] else "wrong"
results_file.write(f"No.{index} {title}\n{pos_result}, {neg_result}, {result}, {actual_result}, {outcome}\n")
correct_results = len(results[results["prediction"] == True])
accuracy = (correct_results / len(results["prediction"]))*100
results_file.write(f"The prediction accuracy is: {accuracy}%")
results_file.close()
def evaluate(self, train_model: pd.DataFrame, test_set: pd.DataFrame, pos_total, neg_total):
results = {
"title": [],
"positive": [],
"negative": [],
"result": [],
"actual_result": [],
"prediction": [],
}
for index, row in test_set.iterrows():
review = row["review"]
pos_prob = pos_total / (pos_total+neg_total)
neg_prob = neg_total / (pos_total+neg_total)
for word in review.split(" "):
if word in train_model["word"].values:
_word = train_model[train_model["word"] == word]
word_pos = _word["positive_prob"].values[0]
word_neg = _word["negative_prob"].values[0]
pos_prob += log10(word_pos)
neg_prob += log10(word_neg)
train_result = "positive" if pos_prob > neg_prob else "negative"
results["title"].append(row["title"])
results["positive"].append(pos_prob)
results["negative"].append(neg_prob)
results["result"].append(train_result)
results["actual_result"].append(row["rating"])
results["prediction"].append(train_result == row["rating"])
return pd.DataFrame(results)
#print(build_vocabulary())