This project focuses on classifying IMDb movie reviews as either positive or negative using linear models. It investigates various text representation methods—including CountVectorizer, TfidfVectorizer, and Word2Vec embeddings—and evaluates their effects on model performance. The project further explores the impact of different regularization strategies (L1 vs. L2) and hyperparameter tuning (e.g., different values of C) to optimize classification accuracy and interpretability. Feature importance analysis and error inspection are also conducted to gain deeper insight into model behavior.