An Algorithm for Friendship Prediction on Location-Based Social Networks

With the development and widely utilization of mobile device, location-based social network (LBSN) has become one important platform for many novel applications. Location information will help discover latent friend relationship and then guide trip, recommend goods and so on. In this paper we propose an algorithm for friendship prediction. Firstly, we adopt the information gain to measure the contribution of different features to human friendship, and extract user social relationship, check-in distance and check-in type as three key features. Secondly, we take the prediction problem as a classification problem and choose SVM (support vector machine) to classify it. At last some experiment results show our algorithm valid to some extent.