Common LMS Log Indicators for Predicting Student Success: A Systematic Literature Review

Abstract

This systematic review synthesizes research on common Learning Management System (LMS) log indicators used to predict student success, addressing challenges of variability and interpretability in leveraging digital behavioral data to forecast academic outcomes. Following PRISMA guidelines, a literature search and screening process using Google Scholar as the primary data source, identified 125 empirical studies for narrative synthesis; no formal meta-analysis was performed, so the ranges reported below reflect the spread of individually reported study findings rather than pooled or averaged effect sizes. The reviewed studies indicate that behavioral indicators such as timeliness of assignment submissions, submission frequency, and regularity of study intervals are repeatedly associated with academic performance, with individual studies reporting correlations as high as r=0.60–0.88, although predictive strength varies considerably by educational context. Tree-based learning models such as Random Forest and Gradient Boosting frequently offer a favorable balance between accuracy (75–96% across studies) and interpretability, while deep learning architectures such as LSTM and GRU are reported to achieve higher accuracy (89–97% in some studies) in capturing non-linear temporal patterns, though these figures are not directly comparable across studies because they were obtained using different evaluation metrics, datasets, and validation procedures. Weekly data aggregation and early engagement metrics (weeks 1–3) are associated with improved timely identification of at-risk students in several studies. Cognitive engagement metrics (e.g., content completion, quiz attempts) are reported to outperform simple frequency indicators in a number of studies, and interpretability techniques such as SHAP values suggest that modifiable behavioral features tend to carry more predictive weight than fixed demographic data. These findings underscore the significant potential of LMS log analysis in understanding student learning behavior. However, the results also highlight methodological differences between studies and limitations in the applicability of the results to other contexts. Therefore, a standardized framework is needed to translate LMS data into applicable, transparent, and equitable educational interventions.