Predictive Modeling of Stress from Sleeping Habits Using Machine Learning Techniques

Stress is a psychological or emotional response triggered by challenging or unavoidable situations, often known as stressors. Understanding human stress levels is essential, as unmanaged stress can lead to adverse outcomes affecting physical health, emotional well-being, and social functioning. Among the many factors influencing stress, sleep patterns play a crucial role, with disruptions often linked to various health complications. This study aims to explore how stress can be effectively identified through machine learning techniques by analyzing sleep-related behaviors. The dataset utilized in this study includes information on sleep patterns along with associated stress levels. To assess the predictive capabilities, six classification models were employed: Multilayer Perceptron (MLP), Random Forest, Support Vector Machine (SVM), Decision Tree, Naïve Bayes, and Logistic Regression. These algorithms were applied to the preprocessed data to evaluate their effectiveness in stress prediction. Experimental results reveal that the Naïve Bayes classifier outperformed other models, achieving an accuracy of 91.27%, along with strong precision, recall, and F-measure scores. It also recorded the lowest values for both Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE). These results indicate that machine learning techniques especially Naive Bayes can serve as reliable methods for evaluating human stress based on sleeping patterns, providing useful insights for early diagnosis and preventive measures.