Product Recommendation Mining Based in User Interested and Rating Prediction on Textual Reviews

Abstract

Any modern internet shopping or social networking plan supports a recommendation mechanism. As an example of outdated recommendation systems, the content - based recommendation system has two key flaws: suggestion redundancy and unpredictability when it comes to new things (cold start). The user's social characteristics, such as personality characteristics and topical interest, may be able to reduce the cold start and eliminate redundant recommendations. Even if the user's history doesn't really contain these or similar items, Meta-Interest predicts the user's interest and the objects linked with these interests. In two ways, the developed scheme is personality-aware: it uses the user's personality features to forecast his or her themes of interest and to link the user's personality facets to the associated items. The proposed system was compared against Recent suggestion systems, such like deep-learning-based recommendation systems and session-based recommendation systems, were compared to the suggested system. In this paper, we present a sentiment-based suggestion strategy (RPS) to increase recommender system prediction accuracy. To enhance predictive performance in recommender systems, we present a sentiment-based rating prediction strategy (RPS) in this paper. To achieve an accurate rating prediction, we combine three criteria into our recommender system: user sentiment resemblance, social sympathetic influence, and item reputation similarity. Sentiment analysis can be carried out at three different levels: review, sentence, and phrase. Analysis at the review level. Our findings suggest that sentiment can accurately describe user preferences, which aids in improving recommendation performance.