Mental health is an important component for overall well-being of the human beings. This study emphasis on examining mental health status based on behavioural analysis and digital device usage patterns. The traditional mental health evaluation methods based on clinical settings often time consuming and self-reported data that can lead to inaccurate prediction. The study uses the Light Gradient Boosting Machine (LightGBM) machine learning algorithm to predict the mental health status by analysing different features, including Social Media Usage Hours, Technology Usage Hours, Gaming Hours and Age. This predicts mental health status that may be Poor, Fair, Good or Excellent. The LightGBM classifier is employed in the proposed study due to its robustness, efficiency and ability to handle large-scale data. This research aims to bridge the gap between technology usage and life style factor by providing a scalable, data-driven solution for mental health prediction. The proposed study uses the SHapley Additive exPlanations (SHAP) values to identify contributions of individual features and combined features for model prediction. The accuracy of the LightGBM algorithm is 84% and precision (82%), recall (81%) and F1 score (80%) highlights the model’s ability to reduce the false predictions and accurately defined the mental health status. The findings can be useful in different sectors, including education, healthcare and workplace wellness programs.