Personalized recommendations based on time-weighted overlapping community detection

• We develop a hybrid recommendation model based on overlapping community detection.
• We propose a temporal overlapping community detection method.
• Temporal factors are incorporated in association rules generation.
• The hybrid model shows better performance in recommendation precision and diversity.

Capturing and understanding user interests are an important part of social media analytics. Users of social media sites often belong to multiple interest communities, and their interests are constantly changing over time. Therefore, modeling and predicting dynamic user interests poses great challenges to providing personalized recommendations in social media analytics research. We propose a novel solution to this research problem by developing a temporal overlapping community detection method based on time-weighted association rule mining. We conducted experiments using MovieLens and Netflix datasets, and our experimental results show that our proposed approach outperforms several existing methods in recommendation precision and diversity.