Examining stable and dynamic factors associated with daily problematic mobile phone use: Evidence from a 14-day diary study using machine learning and network analysis.

This study used machine learning with SHAP to identify key predictors of daily problematic mobile phone use (PMPU), and then applied network analysis to examine the associations between these selected key factors and PMPU across between-person, within-day, and across-day levels. A total of 279 undergraduate students in China completed a baseline assessment and a 14-day daily diary investigation. The results showed that the Random Forest model achieved meaningful out-of-sample prediction of next-day PMPU. SHAP analysis indicated that day-level factors contributed most to prediction (e.g., daily boredom, daily stress, daily loneliness), whereas stable factors showed comparatively smaller contributions. In the between-person network, PMPU was most strongly associated with average daily boredom, while anxiety and stress was the central nodes. In the contemporaneous network, PMPU was most strongly associated with same-day boredom, followed by stress. In the temporal network, prior-day stress, rumination, and boredom emerged as the most prominent lagged predictors of next-day PMPU, and reciprocal paths indicated that higher PMPU predicted increases in next-day stress and boredom. Together, these findings suggest that daily PMPU is shaped primarily by proximal psychological experiences and that boredom and stress constitute core processes in the maintenance of PMPU. The results offer empirical guidance for early identification and for developing targeted prevention and intervention strategies focused on daily stress management and boredom regulation in university students' everyday lives.