A Multi-layered Friend Recommendation System on Twitter

Nowadays, the number of active users on social media is growing. Therefore, the friend recommendation plays a critical role in building a substantial social network. Compared with previous recommendation systems in social networks, our research is not focused on a particular direction (e.g., geographic location, tag) but introduces another field of social media, common interests among users. In the proposed recommendation system on Twitter, the common interests between two users are determined by four features retrieved from a Twitter user account: user introduction, geographic distance between the target user and candidate users, keywords in tweets, and hashtags of tweets. These features are utilized to calculate the similarities between the candidate users and the target user. In the end, a candidate user with a high similarity score is recommended to the target user.