Using predictive analytics to target and improve first year student attrition

Abstract

This article reports on the development and implementation of the Student Success Program, a strategic, University-wide, consistent approach to increasing the success and retention of first year students, during 2014. To this end, a centrally coordinated process integrated historic student, application, enrolment, academic performance and learning management data in a data warehouse. These data were used to build chi-squared automatic iterative detection (CHAID) decision tree models aimed at predicting each student’s risk of attrition. Predictions were made multiple times per year before peak attrition time points to allow for changes in student behaviour and availability of new data. An intervention using peer-to-peer phone-call communication targeted students with the largest predicted risks, to offer support, foster retention and enable a successful outcome in higher education. The accuracy of the chi-squared automatic iterative detection models benefited most from the inclusion of data representing first year student study behaviours.