Enhancing the Performance of Library Book Recommendation System by Employing the Probabilistic-Keyword Model on a Collaborative Filtering Approach

Abstract

This paper proposes the probabilistic-keyword CF method for a library book recommendation system. Our focus is to address the sparsity problem commonly occurs on the Collaborative Filtering (CF) approach. The framework of the method consists of four processes. First, building the circulation and keyword matrices respectively based on the book circulation records and the book keyword attribute data. Second, building the keyword model that takes into account both the book circulation records and the book keyword data. Third, building the probabilistic-keyword model that employs a probabilistic technique to calculate the probability of a user to borrow a book conditional to his/her keyword model. Fourth, generating the top-N book recommendations. Experiment results on a library dataset show that our probabilistic-keyword CF method outperforms the traditional user-based and item-based CF methods in terms of all evaluation metrics. This result conjectures that the probabilistic-keyword CF method that employs the probabilistic-keyword model can enhance the recommendation performance and is able to deal with the sparse dataset better than the traditional methods.