13th Oct 2015, 3:30pm to 5:00pm

Discrimination-Aware Classifiers for Student Performance Prediction & Detecting Students at Risk of Failing

Presented by Ling Luo (University of Sydney) and A/Prof Irena Koprinska (University of Sydney)

UNSW Learning Analytics & Data Science in Education Research Group

Back to meetings

Discrimination-Aware Classifiers for Student Performance Prediction
Presented by Ling Luo (PhD Candidate, University of Sydney)

Mining and using rules that distinguish groups of students based on sensitive attributes such as gender and nationality may lead to discrimination. It is desirable to keep the sensitive attributes during the training of a classifier to avoid information loss but decrease the undesirable correlation between the sensitive attributes and the class attribute when building the classifier. We illustrate, motivate, and solve the problem, and present a case study for predicting student exam performance based on enrolment information and assessment results during the semester. We evaluate the performance of two discrimination-aware classifiers and compare them with their non-discrimination-aware counterparts. The results show that the discrimination-aware classifiers are able to reduce discrimination with trivial loss in accuracy. The proposed method can help teachers to predict student performance accurately without discrimination.

Detecting Students at Risk of Failing
Presented by A/Prof Irena Koprinska (School of Information Rechnologies, University of Sydney)

Detecting students at risk of failing is particularly useful and desirable when it is done in a timely manner and accompanied with practical information that can help with remediation. Our study investigates ways to detect students at risk of failing early in the semester for timely intervention in the context of  a first year computer programming course. We explore whether the use of several student data sources can improve the process: submission steps and outcomes in an automatic marking system that provides instant feedback, student activity in a discussion board and assessment marks during the semester. We built a decision tree classifier that predicts whether students will pass or fail their final exam in the middle of the semester and show that the obtained rules are useful and actionable for teachers and students, and can be used to drive remediation.

If you would like to join this research group or would like more information please contact Lorenzo Vigentini l.vigentini@unsw.edu.au

Event Details
Public Event open to all
40 Seats available


Mathews Building, Room 308

Key Contact
Lorenzo Vigentini