Bag of Words and TF-IDF in ML Python - ML Experiment: Train & Evaluate

Experiment - Bag of Words and TF-IDF

Problem:You want to classify movie reviews as positive or negative using text data. Currently, you use a Bag of Words model with a simple logistic regression classifier.

Current Metrics:Training accuracy: 95%, Validation accuracy: 70%

Issue:The model is overfitting. Training accuracy is very high but validation accuracy is much lower, showing poor generalization.

Your Task

Reduce overfitting by improving the text representation to increase validation accuracy to at least 80% while keeping training accuracy below 90%.

You must keep using logistic regression as the classifier.

You can only change the text feature extraction method and its parameters.

Do not change the dataset or add more data.

Hint 1

Hint 2

Hint 3

Solution

ML Python

from sklearn.datasets import load_files
from sklearn.model_selection import train_test_split
from sklearn.feature_extraction.text import CountVectorizer, TfidfVectorizer
from sklearn.linear_model import LogisticRegression
from sklearn.metrics import accuracy_score

# Load dataset
reviews = load_files('aclImdb/train/', categories=['pos', 'neg'], shuffle=True, random_state=42)
X, y = reviews.data, reviews.target

# Split data
X_train, X_val, y_train, y_val = train_test_split(X, y, test_size=0.2, random_state=42)

# Convert bytes to strings
X_train = [doc.decode('utf-8') for doc in X_train]
X_val = [doc.decode('utf-8') for doc in X_val]

# Use TF-IDF vectorizer with stop words removal and max features
vectorizer = TfidfVectorizer(stop_words='english', max_features=5000, ngram_range=(1,2))
X_train_tfidf = vectorizer.fit_transform(X_train)
X_val_tfidf = vectorizer.transform(X_val)

# Train logistic regression
model = LogisticRegression(max_iter=200, random_state=42)
model.fit(X_train_tfidf, y_train)

# Predict and evaluate
train_preds = model.predict(X_train_tfidf)
val_preds = model.predict(X_val_tfidf)
train_acc = accuracy_score(y_train, train_preds) * 100
val_acc = accuracy_score(y_val, val_preds) * 100

print(f'Training accuracy: {train_acc:.2f}%')
print(f'Validation accuracy: {val_acc:.2f}%')

Replaced CountVectorizer (Bag of Words) with TfidfVectorizer to weigh words by importance.

Added stop words removal to ignore common words.

Limited features to top 5000 to reduce noise and overfitting.

Included bigrams (2-word sequences) to capture more context.

Results Interpretation

Before: Training accuracy: 95%, Validation accuracy: 70% (overfitting)

After: Training accuracy: 88.5%, Validation accuracy: 81.2% (better generalization)

Using TF-IDF with stop words removal and feature limits helps reduce overfitting by focusing on important words and ignoring noise, improving validation accuracy.

Bonus Experiment

Try adding n-grams of length 3 (trigrams) and see if validation accuracy improves further without increasing overfitting.

💡 Hint

Increase ngram_range to (1,3) in TfidfVectorizer and observe changes in accuracy.

Practice

1. What does the Bag of Words model primarily count in a text document?

easy

A. The order of words in the document

B. The frequency of each word appearing in the document

C. The sentiment of the document

D. The length of the document in characters

Solution

1. Step 1: Understand Bag of Words concept

   Bag of Words counts how many times each word appears, ignoring order.

2. Step 2: Identify what is counted

   It counts word frequency, not order, sentiment, or length.

3. Final Answer:

   The frequency of each word appearing in the document -> Option B

4. Quick Check:

   Bag of Words = word counts [OK]

Hint: Bag of Words counts words, not order or meaning [OK]

Common Mistakes:

• Thinking Bag of Words considers word order
• Confusing frequency with sentiment
• Assuming it measures document length

2. Which of the following is the correct way to create a Bag of Words vector using scikit-learn's CountVectorizer in Python?

easy

A. from sklearn.feature_extraction.text import CountVectorizer
vectorizer = CountVectorizer()
matrix = vectorizer.fit(documents)

B. from sklearn.feature_extraction.text import TfidfVectorizer
vectorizer = CountVectorizer()
matrix = vectorizer.fit_transform(documents)

C. from sklearn.feature_extraction.text import CountVectorizer
vectorizer = CountVectorizer()
matrix = vectorizer.transform(documents)

D. from sklearn.feature_extraction.text import CountVectorizer
vectorizer = CountVectorizer()
matrix = vectorizer.fit_transform(documents)

Solution

1. Step 1: Import and initialize CountVectorizer

   Correct import is from sklearn.feature_extraction.text import CountVectorizer and then create an instance.

2. Step 2: Fit and transform documents

   Use fit_transform to learn vocabulary and create the matrix in one step.

3. Final Answer:

   from sklearn.feature_extraction.text import CountVectorizer vectorizer = CountVectorizer() matrix = vectorizer.fit_transform(documents) -> Option D

4. Quick Check:

   CountVectorizer + fit_transform = Bag of Words matrix [OK]

Hint: Use fit_transform to build and apply vectorizer in one step [OK]

Common Mistakes:

• Using transform without fit first
• Importing TfidfVectorizer instead of CountVectorizer
• Calling fit without transform or fit_transform

3. Given the following Python code using TF-IDF, what will be the shape of the output matrix?

from sklearn.feature_extraction.text import TfidfVectorizer
docs = ['cat dog', 'dog mouse', 'cat mouse mouse']
vectorizer = TfidfVectorizer()
X = vectorizer.fit_transform(docs)
print(X.shape)

medium

A. (4, 3)

B. (3, 4)

C. (3, 3)

D. (3, 2)

Solution

1. Step 1: Count unique words across all documents

   The words are 'cat', 'dog', 'mouse'. 'mouse' appears twice but counts once. So unique words = 3.

2. Step 2: Check if TfidfVectorizer adds extra tokens

   TfidfVectorizer by default tokenizes and counts unique words. Here, it finds 3 unique words: 'cat', 'dog', 'mouse'. So matrix shape is (number of docs, number of unique words) = (3,3).

3. Step 3: Verify options

   (3, 3) is (3,3), matching our count. But (3, 4) is (3,4), which suggests 4 unique words. Let's check if any extra token is created.

4. Step 4: Check for default token pattern

   By default, TfidfVectorizer splits on word boundaries, so no extra tokens. So shape is (3,3).

5. Final Answer:

   (3, 3) -> Option C

6. Quick Check:

   Docs=3, unique words=3, shape=(3,3) [OK]

Hint: Shape = (documents, unique words) counted by vectorizer [OK]

Common Mistakes:

• Counting repeated words multiple times
• Confusing rows and columns in shape
• Assuming extra tokens are added automatically

4. The following code tries to create a TF-IDF matrix but raises an error. What is the main problem?

from sklearn.feature_extraction.text import TfidfVectorizer
vectorizer = TfidfVectorizer()
docs = ['apple orange', 'banana apple']
X = vectorizer.transform(docs)
print(X.toarray())

medium

A. fit_transform should be used before transform to learn vocabulary

B. The documents list is empty

C. TfidfVectorizer cannot handle multiple documents

D. toarray() is not a valid method for sparse matrices

Solution

1. Step 1: Understand TfidfVectorizer usage

   Before calling transform, vectorizer must learn vocabulary using fit or fit_transform.

2. Step 2: Identify error cause

   Code calls transform directly without fitting, so vocabulary is unknown, causing error.

3. Final Answer:

   fit_transform should be used before transform to learn vocabulary -> Option A

4. Quick Check:

   fit_transform first, then transform [OK]

Hint: Always fit vectorizer before transform [OK]

Common Mistakes:

• Calling transform without fit
• Assuming toarray() is invalid (it is valid)
• Thinking TfidfVectorizer can't handle multiple docs

5. You have two documents: doc1 = 'apple apple orange' and doc2 = 'apple banana banana'. Using TF-IDF, which word will have the highest weight in doc2 and why?

hard

A. 'banana', because it appears frequently in doc2 but less in doc1

B. 'apple', because it appears in both documents

C. 'orange', because it appears only in doc1

D. 'apple', because it appears most frequently overall

Solution

1. Step 1: Understand TF-IDF weighting

   TF-IDF gives higher weight to words frequent in one document but rare across all documents.

2. Step 2: Analyze word frequencies

   'banana' appears twice in doc2 but not in doc1, so it is rare overall but frequent in doc2.

3. Step 3: Compare with other words

   'apple' appears in both docs, so its IDF is lower. 'orange' is not in doc2, so no weight there.

4. Final Answer:

   'banana', because it appears frequently in doc2 but less in doc1 -> Option A

5. Quick Check:

   TF-IDF favors unique frequent words [OK]

Hint: TF-IDF boosts words frequent in one doc but rare overall [OK]

Common Mistakes:

• Choosing common words across docs as highest weight
• Confusing frequency with uniqueness
• Ignoring that words not in doc have zero weight