Interpersonal Ties and the Social Link Recommendation Problem

Studies on evolving social network is a prevalent topic in today's society. Imagine an individual who tries to enter a social network through socialization. A key concern of the individual involves studying which other members of the social network he or she should interact with, so that the individual may gain the highest positional advantage, i.e., social capital. We ask the question: what ties to make with members of a dynamic social network for an individual who is trying to enter the network, to achieve a better position in the dynamic network? We formalize the dynamic social link recommendation problem. The focus of the problem is placed on modeling the evolution of social networks and the strategies for the target individual to create new ties with others. The result will be a recommendation mechanism that suggests potential members of the network whom the target individual could establish ties with. The goal is not only for the target individual to adopt the recommended output but the recommended output is beneficial to the target individual in terms of gaining a higher positional advantage. The main challenge in building such a recommendation mechanism is to identify potential links that are beneficial while taking into account the network dynamics. We aim to utilize multiple link prediction algorithms and propose edge establishing strategies to solve this specified problem. We evaluate these two sorts of strategies for average recall rate and average closeness centrality metrics, as well as other structural properties such as diameter, distribution of clustering coefficient.