Analyzing Enrolment Patterns: Stacked Ensemble Statistical Learning-Based Approach to Educational Decision Making

AbstractIn Malaysia, Additional Mathematics, equivalent to A-level mathematics, played a vital role in Science, Technology, Engineering, and Mathematics (STEM) education. However, a notable decline in enrolment for the Malaysian Certificate of Education's (SPM’s) Additional Mathematics subject has raised concerns about the implications for Malaysia's STEM workforce and its role in sustainable economic growth. The study’s primary objectives were to identify the determinants that impacted urban upper-secondary students' enrolment in Additional Mathematics within the Kuantan District, Pahang, Malaysia, and to develop a novel stacked ensemble machine learning algorithm based on these determinants, following the CRISP-DM data science methodology. To pursue these objectives, this study collected and analyzed 389 responses from the first-batch urban upper-secondary students in the Kuantan District who had enrolled in the newly revised Standard Based Curriculum for Secondary Schools (KSSM’s) Additional Mathematics syllabus, utilizing a modified research questionnaire and a one-stage cluster sampling technique. The findings revealed that determinants such as education disciplines, ethnicity, gender, mathematics self-efficacy, peer influence, and teacher influence had significantly impacted students' decisions to enroll in Additional Mathematics. Moreover, the introduction of the novel stacked ensemble machine learning algorithm had improved predictive accuracy compared to traditional dichotomous logistic regression algorithms on average, particularly at optimal training-to-test ratios of 70:30, 80:20, and 90:10. These insights were valuable for shaping educational policy and practice, emphasizing the importance of promoting STEM education initiatives and encouraging educators and counselors to empower students to pursue STEM careers while actively promoting gender equality within STEM fields.