Mining Friendships Based on Bipartite Network Though Campus Spatiotemporal Feature

Abstract

The development of the campus network results in a large number of data, which contains student behavior features with implicit spatiotemporal attributes. However, existing mining methods mostly focus on the low-dimensional. It is difficult to cover the dimension of the spatiotemporal attribute. To solve it, based on the bipartite network, this paper proposes a method for mining friendships. Aiming at the feature of the spatiotemporal dataset, a bipartite network is firstly constructed, and divided into sub-networks with the same degree of spatiotemporal nodes. In each sub-network, by using the hypothesis test, edges between co-occurrence nodes of random encounter is deleted. Finally, the friendships network of the students is a projection of the bipartite network. Experiments show that the method can effectively draw the friend relationships between students. Moreover, the friendships network helps to analyze the student behavior, which plays an important role in the decision-making of university.