Gradgroom : an integrated framework for an educational early warning system with mentor matching

Abstract

“Education is the key to success”, one of the most heard motivational statements by all of us. People engage in education at different phases of our lives in different ways. Among them, university education plays a vital role in our academic and professional lives. During university education many undergraduates will be facing several challenges demanding from educational matters to socio-economic problems. In many such situations undergraduates tend to abandon the degree programs halfway leaving them incomplete. Hence creating an Educational Early Warning Systems (EEWS) to predict and identify at-risk undergraduate students in the early stages of the degree programs will improve the graduating ratio against the dropouts. Further, mentoring is another aspect in education where it can be used greatly in the undergraduate studies to speak closer to the students addressing their curiosity and guiding them in the correct direction. There exists many separate frameworks for EEWS and mentoring. Despite an integrated framework that would be capable of identifying students at-risk and then match with the most matched mentor would be impactful and effective to handle by the universities. This study has proposed an integrated framework namely as “GRADGROOM” as a solution to the identified lacuna by extending EEWS framework with mentor matching which perform at-risk undergraduate student prediction and mentor-mentee matching for the identified at-risk undergraduate students. In the study, the proposed framework has been evaluated with two case studies. The framework has compared the use of KNN, SVM and RF algorithms while merging them to a stacking ensemble model with decision tree classifier as the base classifier in at-risk undergraduate student prediction. Further, from the predictions, the framework has extended the work up to logic based mentor-mentee matching and has conducted a successful mentoring program to validate the results. In the study the algorithms KNN, SVM, RF and the stacking ensemble model have given accuracies of 96%, 100%, 92% and 96% respectively proving the success of the prediction at-risk in the EEWS framework. Overall through the case studies and feedback gatherings, study has emphasized the importance of extending the at-risk prediction through EEWS followed with a mentoring process. Moreover, the study has concluded that a proper mentoring process conducted immediately after being identified as at-risk undergraduate students will be highly beneficial to reshape their study patterns to align with the correct route of studying.