Product Recommendation using Feature-Level Analysis

This research study presents a systematic approach to recommend products based on a comprehensive assessment of customer preferences and feedback. The proposed methodology involves a multi-step user interface where customers answer boolean questions to determine their interest, select desired product features, and provide opinions through reviews or ratings. The collected data is organized into a structured dataset, which undergoes preprocessing to address missing values and prepare textual reviews for sentiment analysis using the VADER algorithm. A novel combined metric is derived, integrating sentiment scores and feature ratings, to classify product features into categories (strong, moderate, weak) using K-means clustering. The final recommendation decision is made based on the balance of strong and moderate features relative to the total selected features. This approach offers a systematic framework for personalized product recommendations informed by customer sentiment and feature preferences.