Article

Identifying Behavioral, Sleep, and Digital Predictors of Mental Wellness Across Remote, Hybrid, and In-Person Workers Using Grouped Random Forest Regression

---

Abstract

The rapid expansion of remote and hybrid work arrangements has redefined the behavioral, digital, and physiological determinants of mental wellness. This study aims to identify and compare key predictors of mental well-being across Remote, Hybrid, and In-person workers using a grouped Random Forest regression approach. The ScreenTime vs MentalWellness.csv dataset, comprising 400 anonymized entries, was analyzed after excluding highly collinear variables (stress_level and productivity) and the composite screen_time_hours. Separate Random Forest models were trained for each work mode, with model validity assessed via R², RMSE, and out-of-bag (OOB) scores. Results revealed distinct predictor hierarchies across work arrangements. Sleep quality emerged as the dominant determinant for both Remote (importance = 0.50) and In-person (0.55) workers, while total sleep hours had the strongest effect among Hybrid workers (0.56). Leisure screen time consistently showed a negative influence across all groups, particularly among Remote and In-person employees. Lifestyle factors such as exercise and social interaction contributed moderately to well-being, whereas demographic attributes exerted minimal influence. The In-person model achieved the highest predictive performance (R² = 0.624; RMSE = 12.76), followed by the Hybrid (R² = 0.535) and Remote (R² = 0.126) models. These findings demonstrate that predictors of mental wellness are context-dependent rather than universal. By integrating behavioral, sleep-related, and digital variables into a mode-specific modeling framework, this research provides actionable insights for designing tailored wellness strategies that align with distinct occupational environments in the digital era.

Keywords: mental wellness, work arrangement, sleep quality, screen time, Random Forest regression

How to Cite:

Chantanasut, T., (2025) “Identifying Behavioral, Sleep, and Digital Predictors of Mental Wellness Across Remote, Hybrid, and In-Person Workers Using Grouped Random Forest Regression”, Journal of Digital Society 1(4), 272-286. doi: https://doi.org/10.63913/jds.v1i4.42