Unraveling the Sandman’s Mystery: Predicting and Explaining Sleep Quality Over Eight Years

Sleep quality is a key factor affecting overall health and well-being. This study aims to predict sleep quality the next day using a data-driven machine learning model combined with explainable artificial intelligence (XAI) techniques. We utilized $\mathbf{2, 6 1 1}$ sleep records spanning eight years from a single participant, including behavioral and environmental characteristics. We constructed a regression model with a mean squared error (MSE) of 0.017 and a mean absolute error (MAE) of 0.101, which significantly outperformed the benchmark MAE of 0.11. The analysis showed that factors such as hourly activity, sleep regularity, and urban location were the most critical in predicting sleep quality. The analysis provides insights into the importance of global characteristics and the degree of individual influence, and provides personalized recommendations for optimizing sleep. This study highlights the potential of combining machine learning with XAI to advance personalized sleep management.