Academic burnout detection using behavioral data analysis

Abstract

As the world continues to evolve, educational systems are advancing at an unprecedented pace. However, this rapid change has led to challenges such as academic burnout. Academic burnout is defined as physical, emotional, and mental exhaustion due to academic pressure which leads to symptoms such as reduced motivation, emotional detachment, and lower academic performance. Maslach Burnout Inventory is commonly regarded as the most popular burnout scale, and the student version (MBI-SS) is the instrument of choice in student samples regardless of their fields of study most notably in medicine.Early research has explored the functionality of ML models in predicting burnout by highlighting their effectiveness in analyzing both behavioral and psychological data as well as responses from well-known inventories.Our proposed methodology approches the use of the use of MBI-SS from a different perspective. This paper aims to improve the use of MBI-SS by comparing two of it’s labeling methods, creating a micro-screener which would act as a shorter version of the MBI-SS and using machine learning (ML) models to test the effectiveness of the micro-screener.The models we tested were Logistic Regression (L1/L2), Random Forest (RF), Support Vector Machine (SVM), Radial Basis Function kernel Support Vector Machine (RBF SVM), Extreme Gradient Boosting (XGB), and LGBM (Light Gradient Boosting Machine) and Extra Trees/Extremely Randomized Trees (ET).