With the rise of web services and platforms such as Netflix, Youtube, and Amazon in the last decade, people spend a major time of their day online watching movies, online shopping, advertising, learning and social networking. There is infinite amount of options for users when it comes to choosing the movie they want to watch, or an item they want to purchase which is why recommender systems have taken more and more space in the industry. As a result, researchers and companies always invest on developing recommender systems with better performance. There are different recommender system approaches available in literature, and the user’s requirements including diversity, accuracy, scalability and coverage are key factors to take into consideration. A popular approach to designing recommendation systems is to review the historical data and explore the user- item interactions in order to predict the interest rate of a user for an item. This study includes investigating the most popular collaborative filtering-based methodologies to develop recommendation systems, and it focuses on two highly performed algorithms called Restricted Boltzmann Machines (RBM) and Graph Convolutional Matrix Completion (GC-MC). The performance of these two methodologies with regards to dealing with data sparsity is investigated and conclusions are drawn.