Regresar

Hashtags Recommendations in Twitter based on Collaborative Filtering

Abstract:

Social networks generate each day a vast amount of heterogeneous information, which can be used for recommendation systems, in order to provide personalized service. Currently, these systems are being developed to help Twitter users find items that is of interest to them, as a way to solve the information overload problem. To take advantages provided by collaborative filtering recommendation systems, in this paper we implement the KNN algorithm for the hashtags recommendation in the Twitter social network, both users and hashtags are used to pbkp_redict the missing values in the user-item matrix. As in this matrix the class is binary and imbalanced, we apply a sub-sampling technique to balance the class. A memory-based approach is used to generate the final recommendations. The experiments are carried out on a real dataset, extracted from Twitter social network interactions. The experimental results demonstrate an improvement in the quality of both pbkp_redictions and recommendations using a balanced dataset.