A Better Semantic based Friend Recommendation System for Modern Social Networks

People meet new people on social networks from across the world, eventually bringing the world closer. Existing social systems recommend friends according to person’s social graph which includes mutual friends and social connections, which may not be the case when being friends in real life. A better approach would be to get recommendations according to user’s life-style rather than just social graph. The proposed system provides an intelligent and automated way to predict user’s lifestyle according to his daily activities and interests by taking advantage of sensor-rich smart-phones and recommend friends with high similarity of lifestyles. The user’s data is stored in database and lifestyle is extracted using topic model. By constructing friend-matching graph, our system depicts the similarity of lifestyles between two users. Upon receiving a request, a list of people with highest recommendation scores is returned to the query user. Finally, system integrates a feedback mechanism to further improve the recommendation accuracy.