Novel Personalized Multimedia Recommendation Systems Using Tensor Singular-Value-Decomposition

Publication Date

1-1-2023

Document Type

Conference Proceeding

Publication Title

IEEE International Symposium on Broadband Multimedia Systems and Broadcasting, BMSB

Volume

2023-June

DOI

10.1109/BMSB58369.2023.10211188

Abstract

Nowadays, multimedia data are often produced by various sources, e.g., Internet of Things (IoT), social media, customer databases, etc. As tremendous multimedia data are produced every day, users or E-commerce customers cannot infer information from such data by themselves and therefore recommender systems have been developed to help users to select the products or services which better fit users' preferences or requirements. Nonetheless, there exists few works on the incorporation of side information or multiple attributes about items into the design of a more robust recommender system. In this work, we propose a novel approach based on the third-order tensor singular-value-decomposition (T3-SVD) to design new personalized multimedia recommender systems (PMRSs) for internet users. A PMRS can dynamically adjust its recommendation strategy subject to a particular user's on-line transaction behavior. To evaluate the effectiveness of our proposed PMRS based on T3-SVD, we compare the performances of our proposed new PMRS and two other existing tensor-based recommendation systems over realworld data in terms of normalized root-mean-square error (NRMSE). As a result, our proposed new PMRS greatly outperforms the other two existing systems.

Keywords

E-commerce, normalized root-mean-square error (NRMSE), Personalized multimedia recommendation system (PMRS), tensor singular-value-decomposition

Department

Applied Data Science

Link to Full Text

Share

COinS