Column: product

• Cleaned by:
  • Removing trailing spaces.
  • Stripping extra internal spaces.
  • Eliminating special characters.

Column: feedback (New Column)

• Created by combining the title and reviews_feedback columns.
• This unified text serves as input for the sentiment analysis engine to evaluate the degree of positivity or negativity in customer reviews.

Column: categories

• The original category column was highly inconsistent and nested. The following normalization steps were applied:
  1. Standardization:
     • All category values converted to lowercase.
     • Whitespace trimmed from both ends.
  2. Noise Removal:
     • Identified and removed noisy or irrelevant patterns such as:
       • "retail brand", "walmart", "target", "mazon.co.uk".
  3. Category Unification:
     • Grouped semantically similar but differently named categories using a mapping dictionary.
     • Examples:
       • "phones", "mobile phones", "cellular phones" → phone
       • Similar logic applied for other groups (e.g., tablet, charger, etc.)

Prediction & Output

• Saved results to feature_engineering.csv.