A Novel Approach to Recommendation System Business Workflows: A Case Study for Book E-Commerce Websites

Abstract

Have you ever wondered why a song or a book or a movie becomes so popular that everyone everywhere starts talking about it? If we did not have the technology, we would say that people who love something would start recommending it to their friends and families. We live in the age of technology where there are so many algorithms that can discover the patterns of human interaction and make an excellent guess about someone's opinion about something. These algorithms are building blocks of digital streaming services and E-Commerce websites. These services require as accurate as possible recommendation systems for them to function. While many businesses prefer one type or another of recommendation algorithms, in this study, we developed a hybrid recommendation system for a book E-Commerce website by integrating many popular classical and Deep Neural Network-based recommendation algorithms. Since explicit feedback is unavailable most of the time, all our implementations are on implicit binary feedback. The four algorithms that we were concerned about in this study were the well-known Collaborative filtering algorithms, item-based CF and user-based CF, ALS Matrix Factorization, and Deep Neural Network Based approaches. Consequently, comparing their performances and accuracy, it was not surprising that the Deep Neural Network approach was the most accurate recommender for our E-Commerce website.