Collaborative Filtering and Leaders' Advice Based Recommendation System for Cold Start Users
Chao Xu, Cui Guangcai · 2020
Recommendation system is a specific type of intelligent systems, which exploits historical user ratings on items and auxiliary information to make recommendations on items to the users. It plays a critical role in a wide rang of online shopping, e-commercial services and social networking applications. Collaborative Filtering(CF) is the most popular approaches used for recommendation systems, but it suffers from cold start problem where no rating data is available for new users or items in the system. In this paper, we propose a novel approach to improve cold start problem for new users. The approach is named Leaders Advice Model(LAM). We use it to finish recommendation task. In our approach, we don't make use of private data sources of users, because it is difficult to get and involved in legal problem. Firstly, we use CF approach to generate candidate leaders. Secondly, to identify these candidate leaders, we need to train them. Then, when one new user enters the system, we can interview the user to get his/her interesting trend. Finally, according to new user's interesting trend, our approach can recommend corresponding items to the new user. Some experiments on a MovieLens rating data set of movies are performed, which show that our proposed recommendation approach can improve the cold start problem for new users. The design can be applied to many other recommendation systems for online shopping and social networking applications. The solution of cold start user problem can improve user experience and trust of recommendation systems.