Alleviating Sparsity to Enhance Group Recommendation with Cross-Linked Domain Model

As a new search paradigm emerges, users’ perspectives on information searching shift from finding information to receiving it. Recommender systems (RS) are an emerging method of obtaining information. RS has succeeded in many conventional domains, such as social media and online video platforms like ‘YouTube.’ An insufficient number of user-item ratings triggers obstacles to individual RS and group recommender systems (GRS). Data sparsity becomes an issue as a result of this incompleteness. The accuracy of recommendations given to a group suffers when there is a lack of data within the group. It happens as a result of ineffective group formation, which usually involves individuals who possess sparse user profiles. Most of the research being done now concentrates on this problem after group formation. Assuming that handling data sparsity at the individual level would be more efficient, this study concentrated on data sparsity before the group formation process. The primary goal is to create a cross-domain technique with Linked Open Data (LOD) technology to ensure that problems with data sparsity can be addressed before the group formation process. The experiment of the proposed method leveraging LOD and cross-domain knowledge simultaneously in this study suggests that it achieved the lowest prediction errors compared to other experiments carried out in this study. A more accurate rating prediction can further alleviate data sparsity in the user-item matrix before group formation using user profiles. Hence, this research will enhance the quality of recommendations by reducing data sparsity in user profiles.