Addressing Behavioural Patterns of LateNight Sleepers Using a Supervised Learning Approach

Abstract

Issues related to Sleep have become a big problem among all age Groups, As It leads to long-term physical and mental health problems. This research aims to classify medical conditions according to sleep patterns and lifestyle behaviors associated with them using machine learning algorithms. A dataset was collected that included sleep habits, psychological symptoms, side effects and techniques for managing them. Then, the data was further enriched via SMOTE in order to solve the class imbalance problem. The feature selection algorithm that was used was called Mutual Information. Using Random Forest as a classifier model achieved the highest accuracy of 97.7%, and next is XGBoost at 96.44%, followed closely by Decision Tree with an accuracy of 96.03% and K-Nearest Neighbors with an accuracy of only 83%. Methods that were used also investigated pathways to late sleep and coping mechanisms. It is believed that this research provides an example of the promising scope of supervised machine learning models in health prediction respect of sleep factors and will be useful for people and patients to detect long-term sleep related disorders using data-driven messages.