PREDICTING TIMELY GRADUATION OF POSTGRADUATE STUDENTS USING RANDOM FORESTS ENSEMBLE METHOD

Abstract

Graduation time of students, both undergraduate and postgraduate, has been a prime focus in universities recently. Over the years, there have been numerous research on using data mining techniques to forecast undergrad students' success. However, very few works have been reported on predicting graduation time of postgrads, particularly using data from Nigerian Universities. This research utilized classification techniques using supervised learning to develop a Postgraduate Student Graduation Time Prediction Model (PS_GTPM). Data was collected from Bayero University Kano and the Adaptive synthetic sampling (ADASYN) technique was applied to address the imbalance issue with the data. Then, the model was developed using the Random Forests ensemble technique. From the evaluation results, we found that the data balancing method based on ADASYN technique enhanced the ability of the data mining classifiers to forecast when students will graduate. Also, it was found that the proposed PS_GTPM based on Random Forests Ensemble Method recorded the highest prediction accuracy with more than 83% score compared to the other methods. Largely, PS_GTPM can be used to forecast whether a thesis-based graduate study shall be completed on-time or not.