Predicting Students’ College Drop Out and Departure Decisions by Analyzing their Campus-Based Social Network Text Messages

Abstract

Undergraduate student retention is a key concern in the US higher education system. Having a scientific method for predicting undergraduate student departure would enable institutions to deploy targeted interventions with the goal of retaining a particular student who is at risk of dropping out. We explore the use of Latent Dirichlet Allocation (LDA), Systemic Functional Linguistics (SFL), and new techniques for Social Network Analytics addressing student communications within a novel campus-based closed social networking platform. Our research results indicate that students who were ultimately retained sent three times as many messages than those who were not, and analyzing the patterns of use of the closed social network in an academic setting reliably predicts undergraduate student dropouts and leads to a more effective deployment of retention resources over time.