Exploring Student Profile Features and Their Impact on Learning Performance in Secondary School

Emerging technologies have allowed researchers to easily access educational data, conduct data analysis, and predict students’ learning performance. However, the factors that are essential for the predictive model have not been identified. In the present research, based on the information entropy framework, we firstly identify the factors that influence students’ academic learning performance. Then, we adopt the explainable machine learning frameworks, which are based on logistic regression and support vector machines, to predict student learning achievements. The experiment was conducted on the real-world dataset from the secondary school within two subjects. The results reveal that the feature of the failure records from students’ past performance is a significant factor related to learning achievements. The predictive model based on student profiles achieves up to 86% accuracy for the prediction of learning outcome related to the final grade.