Incremental iterative time spent based ranking model for online activitybased friend-group recommendation systems

Online social networks enhance user experience by connecting users with similar interests. Online friend recommendation is a rapid developing field in data mining. Current social networking services prescribe friends to users in light of their social graphs and mutual friends, which may not be the most proper to reflect a user's taste on friend selection in real lifetime. In this paper propose a system that recommends companions based on the daily activities of users. Here a semantic based friend recommendation is done based on the user's life styles such as posting, chatting, searching, commenting etc. By using text mining technique, we display a user's daily life as life archives, from which his/her ways of life are separated by using the Latent Dirichlet Allocation algorithm. At that point we discover a similarity metric to quantify the similarity of life styles between users as an incremental way, and ascertain user effect as far as ways of life with a similarity matching diagram. Then calculate user impact ranking iterative matrix vector multiplication strategy in user incrementally, so that it would be versatile to vast scale frameworks. Ranking is mainly based on time spent on activities, profile information and feedback factor. At last, we incorporate a feedback component to further improve the proposal precision.