Data mining to generate heuristic rules to calculate early school leaver propensity

School students not completing their education generally have poorer outcomes in life and proportionally become a burden on society. This is a phenomenon of worldwide concern.  In this QUT/DET joint study, the stated client goal was a detailed guide of how to calculate propensity of students becoming a school leaver, based on heuristics rules derived by data mining models built on the student early leaving data. The data mining methods sought to identify groups of students where the proportion of students who became leavers was a measure of propensity; i.e., the propensity to be an early leaver. The outcomes of this study are derived groups with differing propensities of leavers. From the data mining methods utilized, sets of simple human-applicable rules/heuristics of differing complexity were developed for describing these groups. Thus, the client can choose the most appropriate rules for a given application. The set of rules identifies key indicator points in a student’s progress through school, indicating membership of schooling pathways, some with higher risk or propensity of leaving. The heuristic rules can be applied by policy developers or practitioners to identify groups or individuals based on their characteristics in these key areas.