Personalized Recommendations of Products to Users

Abstract

Many organizations utilize recommendation systems to increase their profitability and win over their customers, including Facebook, which suggests friends, LinkedIn, which promotes employment, Spotify, which recommends music, Netflix, which recommends movies, and Amazon, which recommends purchases. When it comes to movie recommendation system, suggestions are made based on user similarities (collaborative filtering) or by considering a specific user's behavior (content-based filtering) that he or she wishes to interact with. Using TF-IDF, cosine similarity method for content-based filtering, and deep learning for a collaborative approach, this study compares two movie recommendation system. The proposed systems are evaluated by calculating the precision and recall values. On a small dataset, a content-based filtering methodology had a precision of 5.6% whereas a collaborative approach had a precision of 57%. Collaborative filtering clearly worked better than content-based filtering. Future improvements involve creating a single hybrid recommendation system that combines a collaborative and content-based approach to improve the outcomes.