Integrated Personalized Book Recommendation using Social Media Analysis

Abstract

Today, most e-commerce sites use product-specific recommendation systems to better user experience. The algorithm used by such sites is - item-toitem collaborative filtering. This matches each user who has purchased and rated items to similar items and then combines those similar items into a recommendation list. The solution proposed in this paper is an integrated book recommendation system that maps the user’s highly rated books with books of a similar genre, maps the interactions of said user on social media to assess the kind of books one is interested in, and considers the collaborative filtering or association mapping between the items. In this model, authors used datasets for the same Goodreads book collection, Amazon and Goodreads reviews, transaction histories, and Twitter data. The proposed solution shall use a weighted measure, k-means clustering, and sentiment analysis. The collaborative filtering will be done using the Apriori mechanism to develop an integrated book recommendation list. The result is a list of 10 books that are recommended for a particular user. The proposed model met 80 percent of the user’s expected recommendations, whereas the simple collaborative model only met 60 percent of the user’s expectations. The collaborative model consisted majority of books by the same author or of a complete contrast genre as the method only considers the choice of other similar users and not similar books. So, the proposed integrated recommendation system is more accurate in its recommendations than a simple collaborative system. This model helps firms recommend the best possible book for book lovers. It also helps book lovers to find the best content as per their interests.