Improving the Scalability of Collaborative Filtering Recommendation with Clustering Techniques
Víctor Botti-Cebriá, Laura Sebastiá, David Monzó, Haritz Garcia · 2023
Nowadays, many online sales platforms make use of recommendation systems to retrieve the most suitable items for the user. Several recommendation techniques can be used, but in many cases, dealing with a huge number of users leads to a significant increase in costs, making them unaffordable to be used in commercial systems. Considering that serving the recommendation to users has to be as fast as possible to get the users' attention and, as far as we are aware, clustering has seen limited adoption in the realm of Deep Learning collaborative filtering techniques, we propose a novel clustering framework recommendation system that builds groups of users using a clustering technique based on users' genre preferences and predicts ratings using Machine Learning and Deep Learning techniques. The aim is to reduce the costs needed to generate the recommendation without decreasing the accuracy of the recommendation. The results will show that the clustering process not only improves the temporal and computational costs of the process but also decreases the error incurred by the recommendation system.