A Cross-Domain Linked Open Data-Enabled in Collaborative Group Recommender System

Abstract

A new search paradigm is continuously evolving, with users' perspectives on information searching shifting from searching for information to receiving information. One of the new methods of receiving information is through recommender systems (RS). RS have proven to be successful in many traditional domains including tourism and books. The group recommender system (GRS) and individual RS challenges are triggered by the limited and incomplete number of user-item ratings. The data sparsity problem emerges because of this incompleteness. Data sparsity in a group has a negative impact on the quality of recommendations made to the group. It occurs due to inefficient group formation, which commonly involves individuals with sparse user profiles. Most current research focuses on this issue after group formation. However, this study concentrated on data sparsity at the individual level prior to the group formation process, with the idea that addressing data sparsity at the individual level would be more efficient. The main goal is to design a cross-domain approach leveraging Linked Open Data (LOD) technology to ensure that data sparsity issues can be addressed before the group formation process is conducted. Thus, by reducing data sparsity in user profiles, this study will benefit in improving the quality of recommendations to the group. Hence, the cross-linked domain model is proposed to be adopted in following process of collaborative GRS. The model designed in 4 phases: (i) proposing an approach of cross-domain with LOD in collaborative GRS, (ii) designing cross-linked domain model, (iii) adopting cross-linked domain model into collaborative GRS, and (iv) performance evaluation.