EUVIMEDEUVIMED entdecken
Crossref

Comparing Early Predictability of Dropout and Underperformance in Online Liberal Arts Courses at a College

Young-bit Moon

Korean Educational Research Association · 31.10.2026 · 64 (6) · 303-324

DOI: 10.30916/kera.64.6.303 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Korean Educational Research Association, online erschienen am 31.10.2026. Verfasst von Young-bit Moon.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Early-warning research using learning management system (LMS) logs has treated academic risk as a single continuum of low achievement. This study separates two phenomena, dropout and underperformance, that leave different traces in the log and compares their predictability from first-half-semester activity. After an audit removed 470 invalid cases, 1,986 records from 11 fully online liberal arts courses at a Korean college (2022-2023) were analyzed. Dropout was failure to sit the midterm, underperformance a total score ratio below 60 among those who sat it. Using weeks 1-7 logs only, random forest was selected from five algorithms and its threshold calibrated within training folds. Of 643 risk cases, 277 (43.1%) were dropouts and 366 (56.9%) underperformers. When attendance was decomposed into attended, late, and absent sessions, dropouts diverged from the non-risk group (absent 3.7 vs. 0.2 of 14; d = .90) while underperformers did not (0.7; d = .29). Under course-grouped validation, dropout prediction reached AUC .873 and average precision .672, 4.8 times the base rate, against .774 and .475, 2.2 times, for underperformance. At a matched recall of .60 the dropout alert flagged 12.7% of students at precision .659, the underperformance alert 33.1% at .430. The contrast held across three validation designs and individual courses, and five attendance variables ranked among the top fifteen predictors of dropout, none for underperformance. LMS logs record whether students participate, not how well they learn; early-warning systems should be designed as dropout alerts.

Zitieren

APA

Moon, Y. (2026). Comparing Early Predictability of Dropout and Underperformance in Online Liberal Arts Courses at a College. Korean Educational Research Association, 64(6), 303-324. https://doi.org/10.30916/kera.64.6.303

Vancouver

Moon Y. Comparing Early Predictability of Dropout and Underperformance in Online Liberal Arts Courses at a College. Korean Educational Research Association. 2026;64(6):303-324. doi:10.30916/kera.64.6.303

BibTeX
@article{Moon2026,
  author = {Moon, Young-bit},
  title = {Comparing Early Predictability of Dropout and Underperformance in Online Liberal Arts Courses at a College},
  journal = {Korean Educational Research Association},
  year = {2026},
  volume = {64},
  number = {6},
  pages = {303--324},
  doi = {10.30916/kera.64.6.303}
}
Kostenlos starten →