Combining Collaborative Filtering and Content Based Filtering for Recommendation Systems

Abstract

Collaborative filtering (CF) and content-based filtering (CBF) are the two primary methods for building Recommender Systems (RS), aiming to predict items that users may like based on their preferences. Each method has its own set of pros and cons, tailored to specific applications. To address limitations and enhance accuracy, the hybrid approach combines both techniques in various ways. In our study, we propose a hybrid recommendation engine to increase precision. While content-based recommendations are based on the similarity of the attributes of the items, collaborative recommendations are based on the similarity of the users. We implement collaborative filtering through deep neural networks, merging it with a content-based model using cosine similarity to forecast item ratings for each user. Our goal is to minimize errors and provide recommendations that align closely with user preferences. We used the MovieLens dataset to evaluate our model, and upon comparison with state-of-the-art methods, we found that our model surpasses the current state-of-the-art.