Machine Learning Analysis of Lifestyle Clusters and Their Association With Mental Health Outcomes in Single Mothers

Abstract

The objective of this study was to identify distinct lifestyle clusters among single mothers in Tehran using machine learning techniques and to examine the association of these clusters with depression, anxiety, and psychological distress. This cross-sectional analytical study was conducted among single mothers residing in Tehran. Data were collected using standardized self-report instruments assessing lifestyle behaviors, including physical activity, diet quality, sleep patterns, screen time, perceived stress, and social support, as well as mental health outcomes. Unsupervised machine learning algorithms were applied to identify latent lifestyle clusters based on standardized lifestyle indicators. Subsequently, supervised machine learning models were used to evaluate the predictive relationship between lifestyle cluster membership and adverse mental health outcomes. Model performance was assessed using cross-validation procedures and standard performance metrics, and feature importance analyses were conducted to enhance interpretability. Unsupervised clustering identified three distinct lifestyle profiles, including a health-oriented cluster, a moderately adaptive cluster, and a high-risk lifestyle cluster. Inferential analyses revealed statistically significant differences in depression, anxiety, and psychological distress across clusters, with the high-risk lifestyle cluster exhibiting the highest levels of all adverse mental health outcomes and the health-oriented cluster showing the lowest levels (p < 0.001). Supervised machine learning models demonstrated good to excellent predictive performance in classifying elevated mental health symptoms, with ensemble-based models achieving the highest accuracy and area under the curve values. Feature importance analyses indicated that perceived stress, sleep quality, and social support were the strongest predictors of adverse mental health outcomes. The findings demonstrate that single mothers’ mental health outcomes are strongly associated with distinct lifestyle patterns and that machine learning approaches offer valuable tools for identifying high-risk profiles and key behavioral targets for intervention.