Improving Personalized Recommendations Through Overlapping Community Detection Using Multi-view Ant Clustering and Association Rule Mining

摘要

Recommender system is a technique to generate meaningful personalized recommendations, suggestions for particular customers. Due to the huge amount of data on the users and their item preferences, the existing recommendation approaches are time-consuming, and they face many performance issues during data processing. Hence, clustering users into overlapping communities will help with the data sparsity problem and enhance recommendation diversity. Another important factor in recommendation system is dynamic, user interest in which the user interest changes over time. Hence, this paper focuses on to develop a multi-view clustering approach using ant clustering method for community detection. To improve the quality of the recommendation, the overlapping communities are further classified based on temporal factors. Finally, for predicting user interest from the communities' adaptive association rule, mining has been applied.