Link prediction in organizational social network based on e-mail communication

Abstract

Last years one can observe a growing interest of researchers and practitioner in organizational social networks based on different forms of relations between employees. Such networks may refer to formal or informal relations between employees. The former of these relations may stem for example from the position of each person in an organization, and the latter may refer to relations between employees based for example on trust or confidence. Social network analysis tools provide several measures which allow to analyze different properties of such networks. The paper focuses on link prediction in organizational social network, where relations between employees are defined by intensity of digital communication (e-mail channel) between them. Different similarity measures have been considered by Authors. Basing on similarity scores calculated for all non-existing links in the network, new potential links with highest likelihood of their formation have been discovered. The results of the experiment (predicted links) have been compared with the results obtained by solving a community detection problem in the same network. They have been also compared with a real structure of the organization investigated in the experiment. The experiment confirmed that the results of solving a link prediction problem in the organizational network may support the process of detecting communities in the network, but it strongly depends on the similarity measure used.