Prediction of the Persistence of Relationships in Social Networks, Considering Previous Reciprocity and Duration

The aim of the paper is to identify permanent or developing relationships between users of the social network and the prediction of survival or ending of such a relationship. We consider a social network with already identified links between users and characteristics of these links. The social network is being analyzed in the longer period of time, with particular attention to the dynamics of development of it as well as relations between users. The dataset collected for the sake of described work is based on the Polish socio-political blog portal salon24.pl. We aggregate attributes of given relations such as: number of interactions, reciprocity, continuity and time of day when the interactions take place. The purpose of this analysis is to build a model that allows estimating the situation that prevails in the network from the point of view of a specific user. The research results allow to determine how important are links with other users, which ones are so strong that they can be considered as permanent, and in which ones it is worth investing in order to improve their position in the social network.