Ranking Social Media News Feeds: A Comparative Study of Personalized and Non-personalized Prediction Models

Home Artificial Intelligence and Its Applications Conference paper Ranking Social Media News Feeds: A Comparative Study of Personalized and Non-personalized Prediction Models Sami Belkacem, Kamel Boukhalfa & Omar Boussaid Conference paper First Online: 12 March 2022 416 Accesses Part of the Lecture Notes in Networks and Systems book series (LNNS,volume 413) Abstract Ranking news feed updates by relevance has been proposed to help social media users catch up with the content they may find interesting. For this matter, a single non-personalized model has been used to predict the relevance for all users. However, as user interests and preferences are different, we believe that using a personalized model for each user is crucial to refine the ranking. In this work, to predict the relevance of news feed updates and improve user experience, we use the random forest algorithm to train and introduce a personalized prediction model for each user. Then, we compare personalized and non-personalized models according to six criteria: (1) the overall prediction performance; (2) the amount of data in the training set; (3) the cold-start problem; (4) the incorporation of user preferences over time; (5) the model fine-tuning; and (6) the personalization of feature importance for users. Experimental results on Twitter show that a single non-personalized model for all users is easy to manage and fine-tune, is less likely to overfit, and it addresses the problem of cold-start and inactive users. On the other hand, the personalized models we introduce allow personalized feature importance, take into consideration the preferences of each user, and allow to track changes in user preferences over time. Furthermore, personalized models give a higher prediction accuracy than non-personalized models.