Data and code for "Course Structure and Early Academic Performance as Predictors of Graduation Qualification"

This record provides the anonymized public dataset and reproducible Jupyter notebooks supporting the manuscript “Course Structure and Early Academic Performance as Predictors of Graduation Qualification: A Multi-major Machine Learning Study”. The repository starts from three public analysis-ready files: course_classification_public.csv, student_model_data_public.csv, and panel_all_major_public.xlsx. It does not include raw student records, personally identifiable information, original grade sheets, admission matching files, or scripts used to construct the public analysis-ready datasets. The notebooks reproduce the main descriptive tables, model comparison results, curriculum credit composition figures, AUC performance plots, SHAP explanations of the selected models, and course-level Top-20 SHAP analyses. The main model dataset includes 1,360 students from four majors: Forestry, Landscape Architecture, Landscape Gardening, and Virtual Landscape Architecture. The course-level panel is used only for supplementary course-level interpretation. The materials are provided for transparency and reproducibility of the published analyses.