Logistic regression for text in NLP - Practice Problems & Coding Challenges

Challenge - 5 Problems

❓ Predict Output

intermediate

What is the output of this logistic regression prediction code?

Given a trained logistic regression model and a text vectorizer, what will be the predicted class for the input text?

NLP

from sklearn.feature_extraction.text import CountVectorizer
from sklearn.linear_model import LogisticRegression

texts = ['good movie', 'bad movie', 'great film', 'terrible film']
labels = [1, 0, 1, 0]

vectorizer = CountVectorizer()
X = vectorizer.fit_transform(texts)

model = LogisticRegression()
model.fit(X, labels)

new_text = ['good film']
X_new = vectorizer.transform(new_text)
prediction = model.predict(X_new)
print(prediction[0])

A 0

B Raises a NotFittedError because model is not trained

C Raises a ValueError due to shape mismatch

D 1

💡 Hint

Think about how the model was trained and what the new text contains.

The model was trained on texts labeled 1 for positive and 0 for negative. 'good film' contains words associated with positive labels, so prediction is 1.

❓ Model Choice

intermediate

Which model is best suited for binary text classification with logistic regression?

You want to classify movie reviews as positive or negative using logistic regression. Which preprocessing and model pipeline is most appropriate?

A Use TfidfVectorizer to convert text to TF-IDF features, then LogisticRegression with L2 regularization

B Use raw text directly as input to LogisticRegression without vectorization

C Use CountVectorizer to convert text to counts, then LogisticRegression with default parameters

D Use PCA on raw text data, then LogisticRegression

💡 Hint

Consider which vectorization method captures word importance better for text.

TfidfVectorizer weights words by importance and LogisticRegression with L2 regularization helps prevent overfitting, making option A best.

❓ Hyperparameter

advanced

Which hyperparameter change will most reduce overfitting in logistic regression for text?

You trained a logistic regression model on text data and see high training accuracy but low test accuracy. Which hyperparameter adjustment is best to reduce overfitting?

A Set solver to 'liblinear'

B Increase the maximum number of iterations

C Increase the regularization strength by decreasing C value

D Use a smaller vocabulary size in the vectorizer

💡 Hint

Think about how regularization controls model complexity.

Decreasing C increases regularization strength, which penalizes large coefficients and reduces overfitting.

❓ Metrics

advanced

Which metric is best to evaluate logistic regression on imbalanced text data?

You have a logistic regression model classifying rare positive events in text. Accuracy is high but misleading. Which metric should you use to better evaluate performance?

A F1-score

B Precision

C Recall

D Accuracy

💡 Hint

Consider a metric that balances false positives and false negatives.

F1-score balances precision and recall, providing a better measure for imbalanced classes than accuracy alone.

🔧 Debug

expert

Why does this logistic regression training code raise a ValueError?

Examine the code below and identify the cause of the ValueError during model training.

NLP

from sklearn.feature_extraction.text import CountVectorizer
from sklearn.linear_model import LogisticRegression

texts = ['happy', 'sad', 'joyful']
labels = [1, 0]

vectorizer = CountVectorizer()
X = vectorizer.fit_transform(texts)

model = LogisticRegression()
model.fit(X, labels)

A LogisticRegression requires dense input, sparse matrix causes error

B Mismatch between number of texts and labels causes ValueError

C CountVectorizer cannot process single-word texts

D Labels must be strings, integers cause error

💡 Hint

Check if the input data and labels have matching lengths.

The labels list has 2 elements but texts list has 3, causing a mismatch error during fit.