Social Sentiment Fusion for Predictive Product Rating System

Abstract

The explosive growth of online product reviews has provided users with abundant information for evaluating products. However, the unstructured nature and volume of reviews often lead to information overload. Traditional recommendation systems primarily rely on structured data such as purchase history or ratings, often neglecting the emotional context present in user-generated content. This paper proposes a sentiment-aware rating prediction model that integrates three critical factors: individual sentiment expression, social influence, and collective product reputation. Using real-world Amazon product reviews, the proposed system demonstrates improved accuracy by analyzing sentiment signals embedded in textual feedback. Evaluation metrics such as precision, recall, F1-score, and RMSE validate the effectiveness of the model. The results confirm that sentiment fusion significantly enhances recommendation quality.