Abstract
Exercise procrastination and exercise addiction reflect two maladaptive patterns at opposite ends of the physical-activity spectrum, yet their socio-ecological correlates among university students remain unclear. We surveyed 570 Chinese university students (mean age = 19.15 ± 1.09 years) and assessed exercise procrastination using the Procrastination in Exercise Scale (PES) and exercise addiction using the Revised Exercise Addiction Inventory (EAI-R). Candidate predictors were drawn from individual characteristics, health behaviors, and interpersonal network factors. To accommodate a relatively large predictor set, we applied LASSO regression with 10-fold cross-validation (λ_min) and evaluated robustness using 1,000 bootstrap resamples and 1,000 repetitions of random 10-fold cross-validation, emphasizing predictors retained in the primary models and consistently selected across both sensitivity analyses. For exercise addiction, the most robust predictors were regular exercise (positively associated) and female (vs. male) sex (negatively associated). For exercise procrastination, later habitual bedtime and more frequent short-video use before sleep were positively associated, whereas regular exercise and a greater number of friends were negatively associated. Regular exercise was the only shared correlate, but it showed opposite directions across outcomes-lower procrastination yet higher addiction-like risk. These findings indicate that campus health promotion may benefit from stratified strategies that both reduce exercise delay and monitor potentially excessive, addiction-like exercise tendencies. Limitations include the cross-sectional, self-reported, single-site design and a coarse personality measure, which restrict causal inference and generalizability; longitudinal replication with improved measurement is warranted.