Optimizing Hybrid Recommendations: VADER-Enhanced Sentiment Analysis
Nossayba Darraz, Ikram Karabila, Anas El-Ansari, Nabil Alami, Mostafa El Mallahi · 2024
Recommendation systems are essential tools for assisting users in discovering relevant and personalized content. The optimization of hybrid recommendations by incorporating sentiment analysis is explored. By analyzing sentiments expressed in user-generated content, such as reviews and feedback, valuable insights into user preferences are gained. Leveraging sentiment analysis improves the effectiveness of recommendations, leading to enhanced user experiences and increased engagement. Our proposed approach utilizes Singular Value Decomposition (SVD) for collaborative filtering, TF-IDF with Lasso for content-based filtering, and VADER for sentiment analysis. We convert continuous sentiment scores into discrete ratings and replace initial ratings with sentiment-derived ones from user reviews. The resulting hybrid system offers improved RMSE and recommendations tailored to user preferences, showcasing the potential of sentiment-aware approaches in enhancing recommendation systems. Additionally, we propose a hybrid recommendation system with NMF and DecisionTreeRegressor-based models, which outperformed all other models with an impressively low RMSE score of 0.092.