A harmonized longitudinal dataset of engineering students for learning analytics and predictive modeling

This article presents a high-dimensional longitudinal dataset (1,656 student records x 56 variables) synthesized from institutional academic administrative records of undergraduate students in an automotive engineering program at an application-oriented university in Vietnam. Unlike static datasets derived from fixed schemas, this resource was extracted from disparate, spreadsheet-based management files across multiple academic years, requiring a systematic process of cleaning, standardization, and harmonization to resolve inconsistencies in nomenclature and coding. The dataset retains real-world curriculum variation across academic years, including differences in course naming, course sequencing, semester placement, and elective pathways. The released data are provided in two parallel tabular formats: a labeled version for immediate interpretation and a harmonized coded version designed for reproducible analytical pipelines. The 56 variables encompass student-level contextual information, harmonized course grades, cumulative performance indicators, scholarship records, and academic status. Beyond descriptive analysis and learning analytics, this repository provides a standardized reusable resource for educational data mining, feature engineering, interpretable machine learning, and predictive modeling studies, including analyses of academic performance and student progression. It also provides a reference dataset for methodological research on data harmonization, the reuse of heterogeneous administrative records, and curriculum analytics.