Predicting Employee Interaction in Large Enterprise Heterogeneous Information Network

Large enterprises find success through the collaboration among the members of the enterprise. It thus becomes a specific business goal to increase collaboration. By thinking of the enterprise as a network one can model interactions over time, and characterize interactions, frequency, and type and marry this to a set of outcomes. Beyond descriptive analytics, one can engage in predictive and prescriptive analytics. In terms of the enterprise social network of a specific company, one prediction task is to identify new interactions between two employees where no interaction has happened previously. We set out to extend the previous work of predicting links in heterogeneous networks to enterprise networks. This paper introduces a modified version of PathPredict by leveraging the hierarchical features of enterprise social networks for predicting future employee interactions. A data set from a company enterprise network called tSpace was extracted and analyzed. We show that the novel measure on hierarchical enterprise paths provides great precision of a co-worker prediction task.