Collaborative Recommendations with Deep Feed-Forward Networks: An Approach to Service Personalization

Abstract

The aim of this article is to discuss an advanced approach to recommendation systems, based on the adoption of Deep Feed-Forward Neural Networks. Recommendation engines are data-driven infrastructures designed to help customers in their decision-making process, and nowadays represent the “state of the art” in designing smart and personalized services, in accordance with the new customer-centric perspective. For this purpose, we followed a quantitative methodological approach, comparing the predictive ability of traditional “Collaborative” recommendation algorithms, like the k-Nearest Neighbors (k-NN) and the Singular Value Decomposition (SVD), with Feed-Forward Neural Networks; given these assumptions, we finally demonstrated that a “Deep” Neural architecture could achieve better results in terms of “loss” generated by the model, laying the foundations for a new, innovative paradigm in service recommendation science.