Machine Learning for prediction of grade
MetadataShow full item record
BI Norwegian Business School (BI), one of Europe’s largest business schools, currently has an estimated budget of 125million NOK on expenses related to exams and grading in 2023. BI has set down a task force with a mission to cut these costs by 15 million NOK. Their report stated that further cost reductions would be possible with measures such as cut printing of digital exams, reduce exam elements for each course and increase the percentage each element has to the total assessment in each course and use digital home exams where possible to mention a few. Beyond this one might argue; the greatest potential for cost reduction lies in the use of AI as machine learning can be a powerful tool. The project objective was initially to investigate whether a machine learning automation model could predict grades similar to a human grader. For several reasons, explained both under Academic Research and under Analysis we looked at what contributions it could give BI, namely a tool for predicting grades, but not replacing a human grader. Rather as a training tool for students. Our best performing model predicted with 90% accuracy if a submission was either (A, B, C) or (D, E, F). Implementing an automation tool for grading, with the right refinements, can be cost-reducing, time saving and a quality consistency improvement opportunity. Why should BI continue to investigate the opportunities that lie in Machine learning for predicting grades? Because the opportunities are nearly endless. Our recommended first step in this direction is harmless, cheap, and useful in addition to allowing for incremental improvements. It could be the first step towards great savings.
Executive Master of Management i Analytics for Strategic Management fra Handelshøyskolen BI, 2023