Stress_Classifier_BERT / stress_categorization_using_bert_transformer.py

Upload stress_categorization_using_bert_transformer.py

c5f4573 verified over 2 years ago

6.68 kB

# -*- coding: utf-8 -*-
"""Stress Categorization Using BERT Transformer.ipynb
Automatically generated by Colaboratory.
Original file is located at
    https://colab.research.google.com/drive/1JZTLCAUBN6XkcQpAWukUsx7dJ5VyC_KR
"""

import numpy as np
import pandas as pd
from tensorflow.keras.preprocessing.text import Tokenizer
from tensorflow.keras.preprocessing.sequence import pad_sequences
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Embedding, Flatten, Dense, LSTM, Dropout
from tensorflow.keras.utils import to_categorical
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import LabelEncoder

!pip install transformers

import pandas as pd
import re
from sklearn.model_selection import train_test_split
from transformers import BertTokenizer, TFBertForSequenceClassification
from transformers import InputExample, InputFeatures
import tensorflow as tf

# 1. Load and inspect the data
data = pd.read_excel('stress_data.xlsx')

# 2. Clean and preprocess the data
def clean_text(text):
    text = text.lower()
    text = re.sub(r'http\S+|www\S+|https\S+', '', text, flags=re.MULTILINE)
    text = re.sub(r'\d+|\W+', ' ', text)
    return text

data['Cleaned_Posts'] = data['Posts'].apply(clean_text)

# Convert string labels to integer indices
label_encoder = LabelEncoder()

data['LabelIndices'] = label_encoder.fit_transform(data['Labels'])

# 3. Tokenize data using BERT's tokenizer
tokenizer = BertTokenizer.from_pretrained("bert-base-uncased", do_lower_case=True)

# Split the data into train and test
train, test = train_test_split(data, test_size=0.2, random_state=42)

# Convert data to InputExample format
def convert_data_to_input_example(data):
    return data.apply(lambda x: InputExample(guid=None, text_a=x['Cleaned_Posts'], text_b=None, label=x['LabelIndices']), axis=1)

train_InputExamples = convert_data_to_input_example(train)
test_InputExamples = convert_data_to_input_example(test)

# Convert to features for BERT input
def convert_input_example_to_feature(example):
    return tokenizer.encode_plus(example.text_a, add_special_tokens=True, max_length=128, pad_to_max_length=True, return_attention_mask=True, return_token_type_ids=False)

train_features = train_InputExamples.apply(convert_input_example_to_feature)
test_features = test_InputExamples.apply(convert_input_example_to_feature)

# Convert features to tensorflow dataset
def convert_features_to_tf_dataset(features, labels):
    def gen():
        for f, l in zip(features, labels):
            yield ({'input_ids': f['input_ids'], 'attention_mask': f['attention_mask']}, l)
    return tf.data.Dataset.from_generator(gen, ({'input_ids': tf.int32, 'attention_mask': tf.int32}, tf.int64), ({'input_ids': tf.TensorShape([None]), 'attention_mask': tf.TensorShape([None])}, tf.TensorShape([])))

def decode_predictions(predictions):
    # Extract predicted indices (assuming predictions is a list of dicts with 'label' keys)
    predicted_indices = [int(pred['label'].split('_')[-1]) for pred in predictions]
    # Decode the indices to original labels
    decoded_labels = label_encoder.inverse_transform(predicted_indices)
    return decoded_labels


train_dataset = convert_features_to_tf_dataset(train_features, train['LabelIndices']).shuffle(100).batch(32).repeat(2)
test_dataset = convert_features_to_tf_dataset(test_features, test['LabelIndices']).batch(32)

# 4. Fine-tune BERT on the dataset
model_new = TFBertForSequenceClassification.from_pretrained("bert-base-uncased", num_labels=len(data['Labels'].unique()))
model_new.compile(optimizer=tf.keras.optimizers.Adam(learning_rate=3e-5, epsilon=1e-08, clipnorm=1.0), loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True), metrics=[tf.keras.metrics.SparseCategoricalAccuracy('accuracy')])
model_new.fit(train_dataset, epochs=1, validation_data=test_dataset)
decode_predictions(model_new.predict("I am financially broken"))

# 5. Evaluate the model
loss, accuracy = model_new.evaluate(test_dataset)
print(f"Test accuracy: {accuracy}")





"""# New Section"""



pip install transformers

import pandas as pd
import re
from sklearn.model_selection import train_test_split
from transformers import BertTokenizer, TFBertForSequenceClassification
from transformers import InputExample, InputFeatures
import tensorflow as tf

# 1. Load and inspect the data
data = pd.read_excel('stress_data.xlsx')

# 2. Clean and preprocess the data
def clean_text(text):
    text = text.lower()
    text = re.sub(r'http\S+|www\S+|https\S+', '', text, flags=re.MULTILINE)
    text = re.sub(r'\d+|\W+', ' ', text)
    return text

data['Cleaned_Posts'] = data['Posts'].apply(clean_text)

# Convert string labels to integer indices
label_encoder = LabelEncoder()

data['LabelIndices'] = label_encoder.fit_transform(data['Labels'])

# 3. Tokenize data using BERT's tokenizer
tokenizer = BertTokenizer.from_pretrained("bert-base-uncased", do_lower_case=True)

tokenizer

data

import pandas as pd
import re
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import LabelEncoder
from transformers import BertTokenizer, TFBertForSequenceClassification
from transformers import InputExample, InputFeatures
import tensorflow as tf

# 1. Load and inspect the data
data = pd.read_excel('stress_data.xlsx')

# 2. Clean and preprocess the data
def clean_text(text):
    text = text.lower()
    text = re.sub(r'http\S+|www\S+|https\S+', '', text, flags=re.MULTILINE)
    text = re.sub(r'\d+|\W+', ' ', text)
    return text

data['Cleaned_Posts'] = data['Posts'].apply(clean_text)

# Convert string labels to integer indices
label_encoder = LabelEncoder()

data['LabelIndices'] = label_encoder.fit_transform(data['Labels'])

data["Labels"]

le_name_mapping = dict(zip(label_encoder.classes_, label_encoder.transform(label_encoder.classes_)))
print(le_name_mapping)

import joblib

# Assuming 'label_encoder' is your LabelEncoder instance
joblib.dump(label_encoder, 'label_encoder.joblib')

label_encoder = joblib.load("/content/label_encoder.joblib")

def decode_predictions(predictions):
    # Extract predicted indices (assuming predictions is a list of dicts with 'label' keys)
    predicted_indices = [int(pred['label'].split('_')[-1]) for pred in predictions]
    # Decode the indices to original labels
    decoded_labels = label_encoder.inverse_transform(predicted_indices)
    return decoded_labels

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-classification", model="NeuEraAI/Stress_Classifier_BERT")

decode_predictions(pipe.predict("I am in huge debts. I have taken huge loans and I can't repay."))

	# -- coding: utf-8 --
	"""Stress Categorization Using BERT Transformer.ipynb

	Automatically generated by Colaboratory.

	Original file is located at
	https://colab.research.google.com/drive/1JZTLCAUBN6XkcQpAWukUsx7dJ5VyC_KR
	"""

	import numpy as np
	import pandas as pd
	from tensorflow.keras.preprocessing.text import Tokenizer
	from tensorflow.keras.preprocessing.sequence import pad_sequences
	from tensorflow.keras.models import Sequential
	from tensorflow.keras.layers import Embedding, Flatten, Dense, LSTM, Dropout
	from tensorflow.keras.utils import to_categorical
	from sklearn.model_selection import train_test_split
	from sklearn.preprocessing import LabelEncoder

	!pip install transformers

	import pandas as pd
	import re
	from sklearn.model_selection import train_test_split
	from transformers import BertTokenizer, TFBertForSequenceClassification
	from transformers import InputExample, InputFeatures
	import tensorflow as tf

	# 1. Load and inspect the data
	data = pd.read_excel('stress_data.xlsx')

	# 2. Clean and preprocess the data
	def clean_text(text):
	text = text.lower()
	text = re.sub(r'http\S+\|www\S+\|https\S+', '', text, flags=re.MULTILINE)
	text = re.sub(r'\d+\|\W+', ' ', text)
	return text

	data['Cleaned_Posts'] = data['Posts'].apply(clean_text)

	# Convert string labels to integer indices
	label_encoder = LabelEncoder()

	data['LabelIndices'] = label_encoder.fit_transform(data['Labels'])

	# 3. Tokenize data using BERT's tokenizer
	tokenizer = BertTokenizer.from_pretrained("bert-base-uncased", do_lower_case=True)

	# Split the data into train and test
	train, test = train_test_split(data, test_size=0.2, random_state=42)

	# Convert data to InputExample format
	def convert_data_to_input_example(data):
	return data.apply(lambda x: InputExample(guid=None, text_a=x['Cleaned_Posts'], text_b=None, label=x['LabelIndices']), axis=1)

	train_InputExamples = convert_data_to_input_example(train)
	test_InputExamples = convert_data_to_input_example(test)

	# Convert to features for BERT input
	def convert_input_example_to_feature(example):
	return tokenizer.encode_plus(example.text_a, add_special_tokens=True, max_length=128, pad_to_max_length=True, return_attention_mask=True, return_token_type_ids=False)

	train_features = train_InputExamples.apply(convert_input_example_to_feature)
	test_features = test_InputExamples.apply(convert_input_example_to_feature)

	# Convert features to tensorflow dataset
	def convert_features_to_tf_dataset(features, labels):
	def gen():
	for f, l in zip(features, labels):
	yield ({'input_ids': f['input_ids'], 'attention_mask': f['attention_mask']}, l)
	return tf.data.Dataset.from_generator(gen, ({'input_ids': tf.int32, 'attention_mask': tf.int32}, tf.int64), ({'input_ids': tf.TensorShape([None]), 'attention_mask': tf.TensorShape([None])}, tf.TensorShape([])))

	def decode_predictions(predictions):
	# Extract predicted indices (assuming predictions is a list of dicts with 'label' keys)
	predicted_indices = [int(pred['label'].split('_')[-1]) for pred in predictions]
	# Decode the indices to original labels
	decoded_labels = label_encoder.inverse_transform(predicted_indices)
	return decoded_labels


	train_dataset = convert_features_to_tf_dataset(train_features, train['LabelIndices']).shuffle(100).batch(32).repeat(2)
	test_dataset = convert_features_to_tf_dataset(test_features, test['LabelIndices']).batch(32)

	# 4. Fine-tune BERT on the dataset
	model_new = TFBertForSequenceClassification.from_pretrained("bert-base-uncased", num_labels=len(data['Labels'].unique()))
	model_new.compile(optimizer=tf.keras.optimizers.Adam(learning_rate=3e-5, epsilon=1e-08, clipnorm=1.0), loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True), metrics=[tf.keras.metrics.SparseCategoricalAccuracy('accuracy')])
	model_new.fit(train_dataset, epochs=1, validation_data=test_dataset)
	decode_predictions(model_new.predict("I am financially broken"))

	# 5. Evaluate the model
	loss, accuracy = model_new.evaluate(test_dataset)
	print(f"Test accuracy: {accuracy}")





	"""# New Section"""



	pip install transformers

	import pandas as pd
	import re
	from sklearn.model_selection import train_test_split
	from transformers import BertTokenizer, TFBertForSequenceClassification
	from transformers import InputExample, InputFeatures
	import tensorflow as tf

	# 1. Load and inspect the data
	data = pd.read_excel('stress_data.xlsx')

	# 2. Clean and preprocess the data
	def clean_text(text):
	text = text.lower()
	text = re.sub(r'http\S+\|www\S+\|https\S+', '', text, flags=re.MULTILINE)
	text = re.sub(r'\d+\|\W+', ' ', text)
	return text

	data['Cleaned_Posts'] = data['Posts'].apply(clean_text)

	# Convert string labels to integer indices
	label_encoder = LabelEncoder()

	data['LabelIndices'] = label_encoder.fit_transform(data['Labels'])

	# 3. Tokenize data using BERT's tokenizer
	tokenizer = BertTokenizer.from_pretrained("bert-base-uncased", do_lower_case=True)

	tokenizer

	data

	import pandas as pd
	import re
	from sklearn.model_selection import train_test_split
	from sklearn.preprocessing import LabelEncoder
	from transformers import BertTokenizer, TFBertForSequenceClassification
	from transformers import InputExample, InputFeatures
	import tensorflow as tf

	# 1. Load and inspect the data
	data = pd.read_excel('stress_data.xlsx')

	# 2. Clean and preprocess the data
	def clean_text(text):
	text = text.lower()
	text = re.sub(r'http\S+\|www\S+\|https\S+', '', text, flags=re.MULTILINE)
	text = re.sub(r'\d+\|\W+', ' ', text)
	return text

	data['Cleaned_Posts'] = data['Posts'].apply(clean_text)

	# Convert string labels to integer indices
	label_encoder = LabelEncoder()

	data['LabelIndices'] = label_encoder.fit_transform(data['Labels'])

	data["Labels"]

	le_name_mapping = dict(zip(label_encoder.classes_, label_encoder.transform(label_encoder.classes_)))
	print(le_name_mapping)

	import joblib

	# Assuming 'label_encoder' is your LabelEncoder instance
	joblib.dump(label_encoder, 'label_encoder.joblib')

	label_encoder = joblib.load("/content/label_encoder.joblib")

	def decode_predictions(predictions):
	# Extract predicted indices (assuming predictions is a list of dicts with 'label' keys)
	predicted_indices = [int(pred['label'].split('_')[-1]) for pred in predictions]
	# Decode the indices to original labels
	decoded_labels = label_encoder.inverse_transform(predicted_indices)
	return decoded_labels

	# Use a pipeline as a high-level helper
	from transformers import pipeline

	pipe = pipeline("text-classification", model="NeuEraAI/Stress_Classifier_BERT")

	decode_predictions(pipe.predict("I am in huge debts. I have taken huge loans and I can't repay."))