Overview

This module prepares product review data for sentiment analysis by cleaning key columns, generating features, and training a Logistic Regression model using TF-IDF.

Product Column Cleanup

• Removed non-alphanumeric characters.
• Trimmed and normalized whitespace.

Feedback Column Creation

• Combined title and reviews feedback into a new feedback column.
• Handled missing values gracefully.

Normalized Categories

• Lowercased and trimmed all category entries.
• Removed noise (e.g., "walmart", "retail brand").
• Mapped similar terms to a unified label (e.g., "mobile phones" → "phones").

Sentiment Labeling

• Ratings ≥ 4 → Positive (1)
• Ratings < 4 → Negative (-1)