Personalized news recommender system using deep reinforcement learning

Abstract

News recommender system recommends news based on the user's interests and the content of the news articles. News recommender systems are different from other recommender systems as it always has to recommend latest news articles, essentially making it a item cold start problem. Also the system should take into consideration the changing user preferences. The ratings in such a system is collected as implicit feedback i.e. the user will click on the news item if the recommendation looks interesting to the user. hough we have methods like collaborative filtering, Content Based model, knowledge based methods they are ill equipped to handle the item cold start problem and also they do not capture the changing user preferences. Though techniques using deep learning have been used to solve these problems they suffer from over specialization and do not consider the changing user preferences. In this thesis we propose the use of deep reinforcement learning as a means to recommend news articles by taking into consideration the preferences of user and also adapt as per the change in user preferences. Along with this it also maintains the diversity of the news articles that are recommended to the users.