A Data-driven Paradigm Shift: Predicting Student Academic Performance: a Machine Learning Approach Revealing the Primacy of Personality over Learning Styles

Academic performance is a debated research topic in the educational context. Many researchers have proposed different frameworks and set of factors that enhance the student's academic performance. Demographic and academic-related factors are the most recognized determinants of the target outcome. On the other hand, there is an ongoing debate about which contributes more to predicting student performance among the learning styles and stable personality traits. This study investigated this using a machine learning predictive approach on a student dataset, including learning styles, demographic factors, academic subject preferences, and personality traits. The cross-sectional study design is used to collect the data from 636 university undergraduate students. Random forest and support vector machine models are used in conjunction with the synthetic minority over-sampling technique to address the class imbalance issue. The random forest model with the SMOTE approach achieved better performance, with an 85% accuracy in validating a data-driven approach for this problem. The results also provided an interesting insight that stable personality traits influence student performance more than learning styles, as feature importance analysis showed that conscientiousness is the most important feature, followed by extraversion. However, this study proposed a data-driven shift towards individual personality support systems that highlight personality-related factors as important contributors to students' academic success. This study also recommends some future work for the researchers to collect the data more evenly from all the students' grade classes and add more personality-related factors to validate the proposed model's performance further.