Stress is a natural physiological response to daily challenges, yet when it becomes chronic or excessive, it can disrupt bodily functions, contribute to mental health disorders, and significantly affect quality of life. Accurately predicting stress levels is thus essential for early intervention and stress management. This study explores the potential of Deep Neural Networks (DNNs) to predict daily stress levels based on sleep-derived data. While multiple machine learning models were evaluated—including XGBoost, Random Forest (RF), Support Vector Machine (SVM), and Artificial Neural Network (ANN) the DNN model demonstrated superior performance across all evaluation metrics. Accuracy, confusion matrix, classification report, and ROC curves consistently highlighted the DNN’s enhanced ability to discriminate between stress levels. Our findings underscore the effectiveness of DNNs in capturing complex, non-linear patterns in physiological data, positioning them as a powerful tool for stress prediction and personalized mental health monitoring.