A data mining tool for predicting student withdrawal

The research aims to develop a data mining tool for predicting student withdrawal based on student data available within Swinburne University. The following five research questions were answered during the research: (i) Which students are most likely to withdraw from their courses? (ii) What are the appropriate variables for predicting student course withdrawal? (iii) What is the most robust and easy-to-understand data mining model for predicting student withdrawal? (iv) What is the best approach to implement the data mining model for ease of use and understanding of the model by non-expert users? (v) How to make the data mining model appropriate and relevant in the long term? Models for predicting student withdrawal have been developed: the Initial and the Final Models. The Initial Model was applied to undergraduate and postgraduate students, who have studied at university at least for one semester (continuing students). However, a business evaluation of the model has shifted focus from all continuing students to commencing and continuing undergraduate students. That is why Final Models were developed for predicting commencing and continuing undergraduate student withdrawal. Additionally, after a business evaluation, student data from the online student management system (Blackboard Analytics) became available for use in the Final Models.