Educational Profile Data and Digital Footprint from LMS for the Analysis of Academic Retake Outcomes

AbstractThe dataset comprises educational profiles and LMS digital footprints of 18,192 university students who failed to pass the examination session in the preceding semester. The data spans three consecutive semesters: two prior semesters and the current semester. It provides a unique opportunity for longitudinal learning analytics, survival analysis, and predictive modeling of academic retake outcomes. MotivationWhile many open educational datasets focus on static snapshots of student performance, this dataset captures the stochastic dynamics of at-risk students. It includes weekly temporal resolution of LMS interactions (weeks 3, 4, 7, 10, 17, 18) standardized by study group averages, allowing researchers to model behavioral trajectories, identify latent states of engagement, and apply time-series analysis. Furthermore, it specifically targets the cohort of students with academic debts, a critical group for studying dropout prevention and intervention strategies. Data Structure Overview The dataset includes the following feature groups: Demographics & Academic Profile: Student ID, academic year, year of study, gender, age (normalized), international status, faculty, funding type, and degree level. Academic History & Debts: Number of academic leaves, transfers, prior dismissals, and historical GPA. Retake Dynamics (Previous & 2 Semesters Prior): Comprehensive metrics on courses enrolled, pass/fail and graded exams, initial GPA, and GPA after first and second retakes. Includes counts of outstanding debts at each stage. Proportional Metrics: Ratios of failed courses to total courses before and after retakes. LMS Digital Footprint (Temporal): Weekly standardized metrics for active clicks, effective clicks (interactions altering course content), average e-course scores, and downtime (inactivity duration in weeks) across weeks N = 3, 4, 7, 10, 17, 18 for the current, previous, and two-semesters-prior periods. Citation and Related Work If you use this dataset in your research, please cite the accompanying original article: Esin, R. V., & Kustitskaya, T. A. (2026). Two-Level Monitoring System for Preventing Academic Failure, Based on Predictive Models and SHAP Analysis. Education Sciences, 16(6), 842. https://doi.org/10.3390/educsci16060842