Predictive Analytics for Tertiary Learners in New Zealand Who Are at Risk of Dropping Out of Education

Abstract

This industry showcase covers a proof-of-concept predictive model in the education sector of New Zealand. Jade Software worked with New Zealand’s Tertiary Education Commission on research to find out how to predict the likelihood of learners dropping out. Our model informs the implementation of intervention programs to support learners in completing their qualifications. The goal of this research is to identify a common data set across multiple types of tertiary education organizations and develop predictive models using the data set. We found that the Single Data Return is a viable data source to form a base model. By comparing the area under the receiver operator characteristic curve, we show that additional data sources, including the attendance data and the learner’s results, are helpful in improving model performance. We also developed an interactive dashboard to facilitate estimating the return on investment for intervention programs and the optimal intervention threshold.