Student course grade prediction using the random forest algorithm: Analysis of predictors' importance

Background: Universities need to find strategies for improving student retention rates. Predicting student academic performance enables institutions to identify underachievers and take appropriate actions to increase student completion and lower dropout rates. Method: In this work, we proposed a model based on random forest methodology to predict students' course performance using seven input predictors and find their relative importance in determining the course grade. Seven predictors were derived from transcripts and recorded data from 650 undergraduate computing students. Results: Our findings indicate that grade point average and high school score were the two most significant predictors of a course grade. The course category and class attendance percentage have equal importance. Course delivery mode does not have a significant effect. Conclusion: Our findings show that courses students at risk find challenging can be identified, and appropriate actions, procedures, and policies can be taken.