Social Network Friend Recommendation System Using Semantic Web

In today's growing era of social networking services friend recommendations may not be most relevant to reflect users expectations on friend selection in real life because friend recommendations are based on social graphs which uses tastes and peoples are the basis for recommendations. In this paper, a personalised friendbook recommendation mobile application is presented, which is a novel semantic based friend recommendation system for social networking services. It recommends friend to system users based on their life styles, habits, locations or user profiles. This system measures the similarities of life styles, habits, locations or user profiles between users and recommends friends if their life styles, habits, locations or user profiles have higher similarities. Using text mining, daily activities of users are modelled as a life documents. This life document is used for extracting user's life styles or habits by using Latent Dirichlet Allocation algorithm. The similarity metric is proposed to measure similarity of life styles or habits between users and impact of life styles or habits is calculated using friend matching graph. Friendbook returns a list of users with highest recommendation score according to users request. After receiving recommendations user can provide feedback through feedback mechanism which is used to improve further recommendation accuracy. Finally, we implemented this system using Java, J2EE technology and android development kit to provide personalised friend recommendations.