The Implementation of Hybrid Semantic Ontology-based Model on Movie Recommendation System

Abstract

This paper adopts the Hybrid Semantic Ontology-based (HSO) model for a movie recommendation system. HSO consists of Collaborative Filtering (CF) and Content-based (CB) modules that respectively implement Matrix Factorization (MF) and ONTO Semantic Similarity algorithms. Since the feedback data type influences the MF algorithm choice, we individually implement the Non-Negative Matrix Factorization (NMF) and Singular Value Decomposition (SVD) algorithms for handling the movie rating data. Accordingly, our proposed methods are called HSO-NMF and HSO-SVD. Meanwhile, since the domain determines the ontology, we build and use a new movie ontology on the CB module. The experiments show that HSO performs the best when implemented using the SVD algorithm. On average, the increased percentages of HSO-SVD to HSO-NMF are 1.18% and 1.62% in Precision and NDCG metrics, respectively. The experiments also show that implementing the Hybrid model yields more accurate results than the CB or CF model.