Skip to content

add tf-idf-clf #44

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 1 commit into
base: main
Choose a base branch
from
Open

add tf-idf-clf #44

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
174 changes: 174 additions & 0 deletions examples/tf-idf-clf/rel-movielens1m/classification/train.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,174 @@
# tf-idf-clf
# Predicting Movie Genres Based on Plot Summaries
# https://arxiv.org/pdf/1801.04813.pdf
# micro: 0.32867132867132864; macro: 0.13443582645637167
# 0.4622201919555664s
# N/A
# python train.py

import time
import pandas as pd
import nltk
import re
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.linear_model import LogisticRegression
from sklearn.multiclass import OneVsRestClassifier
from sklearn.metrics import f1_score
from sklearn.preprocessing import MultiLabelBinarizer
nltk.download('stopwords')
from nltk.corpus import stopwords
stop_words = set(stopwords.words('english'))

t_begin = time.time()

# extract useful columns from given tables
col = [0, 1, 3, 9]
train = pd.read_csv("../../../rllm/datasets/rel-movielens1m/classification/movies/train.csv", usecols=col)
val = pd.read_csv("../../../rllm/datasets/rel-movielens1m/classification/movies/validation.csv", usecols=col)
test = pd.read_csv("../../../rllm/datasets/rel-movielens1m/classification/movies/test.csv", usecols=col)

# visualization
# def freq_words(x, terms=30):
# all_words = ' '.join([text for text in x])
# all_words = all_words.split()
# fdist = nltk.FreqDist(all_words)
# words_df = pd.DataFrame({'word':list(fdist.keys()), 'count':list(fdist.values())})

# d = words_df.nlargest(columns="count", n=terms)

# plt.figure(figsize=(12,15))
# ax = sns.barplot(data=d, x="count", y="word")
# ax.set(ylabel='Word')
# plt.show()


# remove redundant information from plot
def clean_text(text):
text = re.sub("\'", "", text)
text = re.sub("[^a-zA-Z]"," ", text)
text = ' '.join(text.split())
text = text.lower()
return text


# remove meaning words in plot
def remove_stopwords(text):
no_stopword_text = [w for w in text.split() if not w in stop_words]
return ' '.join(no_stopword_text)


def train_test():
# train
# visualization
# train.head()
# extract data from 'Genre'
genres_train = []
for i in train['Genre']:
genres_train.append(i.split("|"))

train['Genre_'] = genres_train
# train.head()

genres_val = []
for i in val['Genre']:
genres_val.append(i.split("|"))

val['Genre_'] = genres_val
# val.head()

# visualization
# all_genres = sum(genres_train,[])
# print(len(set(all_genres)))

# all_genres = nltk.FreqDist(all_genres)

# all_genres_df = pd.DataFrame({'Genre': list(all_genres.keys()),
# 'Count': list(all_genres.values())})

# g = all_genres_df.nlargest(columns="Count", n = 50)
# plt.figure(figsize=(12,15))
# ax = sns.barplot(data=g, x= "Count", y = "Genre")
# ax.set(ylabel = 'Count')
# plt.show()

# clean plot data
train['cleanPlot'] = train['Plot'].apply(lambda x: clean_text(str(x)))
val['cleanPlot'] = val['Plot'].apply(lambda x: clean_text(str(x)))

# visualization
# freq_words(train['cleanPlot'], 100)

# delete meaningless words in the plot
train['cleanPlot'] = train['cleanPlot'].apply(lambda x: remove_stopwords(x))
val['cleanPlot'] = val['cleanPlot'].apply(lambda x: remove_stopwords(x))

# visualization
# freq_words(train['cleanPlot'], 100)

# label collection
multilabel_binarizer = MultiLabelBinarizer()
y = multilabel_binarizer.fit_transform(train['Genre_'])

tfidf_vectorizer = TfidfVectorizer(max_df=0.8, max_features=10000)

x_train = train['cleanPlot'].astype(str)
x_val = val['cleanPlot'].astype(str)

# apply tf-idf to construct vectorized space
x_train_tfidf = tfidf_vectorizer.fit_transform(x_train)
x_val_tfidf = tfidf_vectorizer.transform(x_val)

# load model
lr = LogisticRegression()
clf = OneVsRestClassifier(lr)

# train and valid
clf.fit(x_train_tfidf, y)
y_val = multilabel_binarizer.fit_transform(val['Genre_'])
clf.fit(x_val_tfidf, y_val)

# visualization
# y_pred = clf.predict(x_val_tfidf)
# multilabel_binarizer.inverse_transform(y_pred)[1]

# predict on the validation set to fine-tune hyperparameters
y_pred_prob = clf.predict_proba(x_val_tfidf)
t = 0.1
y_pred_new = (y_pred_prob >= t).astype(int)

# validation result
f1_score(y_val, y_pred_new, average='micro', zero_division=1)
f1_score(y_val, y_pred_new, average='macro', zero_division=1)

# test
# extract data from 'Genre'
genres_test = []

for i in test['Genre']:
genres_test.append(i.split("|"))

test['Genre_'] = genres_test

# clean plot data
test['cleanPlot'] = test['Plot'].apply(lambda x: clean_text(str(x)))
test['cleanPlot'] = test['cleanPlot'].apply(lambda x: remove_stopwords(x))

y_test = multilabel_binarizer.fit_transform(test['Genre_'])
x_test = test['cleanPlot'].astype(str)

x_test_tfidf = tfidf_vectorizer.transform(x_test)

# predict
y_pred_prob = clf.predict_proba(x_test_tfidf)
t = 0.095
q_pred = (y_pred_prob >= t).astype(int)

print("micro: ", f1_score(y_test, q_pred, average="micro"))
print("macro: ", f1_score(y_test, q_pred, average="macro"))


if __name__ == "__main__":
# train & test
train_test()
t_end = time.time()
print("Total time:", t_end - t_begin)
136 changes: 136 additions & 0 deletions examples/tf-idf-clf/wiki/classification/train.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,136 @@
# tf-idf-clf
# Predicting Movie Genres Based on Plot Summaries
# https://arxiv.org/pdf/1801.04813.pdf
# micro: 0.3643202579258463; macro: 0.06953107551293945
# 12.427888631820679s
# N/A
# python train.py

import time
import pandas as pd
import nltk
import re
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.linear_model import LogisticRegression
from sklearn.multiclass import OneVsRestClassifier
from sklearn.metrics import f1_score
from sklearn.preprocessing import MultiLabelBinarizer
from sklearn.model_selection import train_test_split
nltk.download('stopwords')
from nltk.corpus import stopwords
stop_words = set(stopwords.words('english'))

t_begin = time.time()

# extract useful columns from given tables
col = [0, 2, 6]
meta = pd.read_csv("../dataset/wiki_dataset.csv", usecols=col)

# visualization
# def freq_words(x, terms=30):
# all_words = ' '.join([text for text in x])
# all_words = all_words.split()
# fdist = nltk.FreqDist(all_words)
# words_df = pd.DataFrame({'word':list(fdist.keys()), 'count':list(fdist.values())})

# d = words_df.nlargest(columns="count", n=terms)

# plt.figure(figsize=(12,15))
# ax = sns.barplot(data=d, x="count", y="word")
# ax.set(ylabel='Word')
# plt.show()


# remove redundant information from plot
def clean_text(text):
text = re.sub("\'", "", text)
text = re.sub("[^a-zA-Z]"," ", text)
text = ' '.join(text.split())
text = text.lower()
return text


# remove meaning words in plot
def remove_stopwords(text):
no_stopword_text = [w for w in text.split() if w not in stop_words]
return ' '.join(no_stopword_text)


def train_test():
# train
# visualization
# meta.head()
# extract data from 'Genre'
genres = []
for i in meta['Genre']:
genres.append(i.split("|"))

meta['Genre_'] = genres
# meta.head()

# visualization
# all_genres = sum(genres,[])
# print(len(set(all_genres)))

# all_genres = nltk.FreqDist(all_genres)

# all_genres_df = pd.DataFrame({'Genre': list(all_genres.keys()),
# 'Count': list(all_genres.values())})

# g = all_genres_df.nlargest(columns="Count", n = 50)
# plt.figure(figsize=(12,15))
# ax = sns.barplot(data=g, x= "Count", y = "Genre")
# ax.set(ylabel = 'Count')
# plt.show()

# clean plot data
meta['cleanPlot'] = meta['Plot'].apply(lambda x: clean_text(str(x)))

# visualization
# freq_words(meta['cleanPlot'], 100)

# delete meaningless words in the plot
meta['cleanPlot'] = meta['cleanPlot'].apply(lambda x: remove_stopwords(x))

# visualization
# freq_words(meta['cleanPlot'], 100)

# label collection
multilabel_binarizer = MultiLabelBinarizer()
y = multilabel_binarizer.fit_transform(meta['Genre_'])

tfidf_vectorizer = TfidfVectorizer(max_df=0.8, max_features=10000)

x_train, x_test, y_train, y_test = train_test_split(meta['cleanPlot'], y, test_size=0.2, random_state=9)
# x_train, x_test, y_train, y_test = train_test_split(meta['cleanPlot'], y, test_size=0.2, stratify=meta['Genre_'])

# apply tf-idf to construct vectorized space
x_train_tfidf = tfidf_vectorizer.fit_transform(x_train)
x_test_tfidf = tfidf_vectorizer.transform(x_test)

# load model
lr = LogisticRegression()
clf = OneVsRestClassifier(lr)

# train and valid
clf.fit(x_train_tfidf, y_train)

# visualization
# y_pred = clf.predict(x_test_tfidf)
# multilabel_binarizer.inverse_transform(y_pred)[1]

# test
# predict
y_pred_prob = clf.predict_proba(x_test_tfidf)
t = 0.4
q_pred = (y_pred_prob >= t).astype(int)

print("micro: ", f1_score(y_test, q_pred, average="micro"))
print("macro: ", f1_score(y_test, q_pred, average="macro"))


if __name__ == "__main__":
# train & test
train_test()
t_end = time.time()
print("Total time:", t_end - t_begin)
Loading