A Study of Factors Influencing College Students’ Online Cramming for Finals: An Emerging Norm in the IT-Enabled Learning Context with Random Forest Analysis

With the ongoing digitalization of higher education, college students’ learning behaviors are rapidly shifting. This study, based on 532 valid questionnaires, examines last-minute online cramming before final exams through four dimensions—time pressure, perceived platform usefulness, peer influence, and course experience—and aims to identify their joint effects on cramming intention. To complement traditional analyses, a Random Forest model is employed to capture nonlinear and interaction effects, and SHAP values are used to provide interpretable insights into each factor’s contribution. The results show that: (1) online cramming has become a widespread strategy, reflecting issues related to time management and motivation; (2) students’ perceptions of platform usefulness vary substantially, indicating that online resources require quality enhancement; (3) peer influence peaks during the sophomore year; (4) course experience affects cramming mainly in an indirect manner; and (5) gender and major show no significant differences, suggesting the behavior is common across groups. Machine-learning findings further highlight time pressure as the most influential predictor, exhibiting a notable nonlinear threshold. These findings suggest that universities should strengthen time-management training, improve platform resources, emphasize continuous assessment, and foster supportive peer study networks.