A Hybrid and Ensemble Deep Learning Approach for Prediction and Analysis of Sleep Quality using Wearable IoT Device Data for Improved Accuracy

Abstract

Sleep quality refers to how well a person sleeps during the night. There are many factors that can affect sleep quality, including stress, anxiety, diet, exercise, and environmental factors such as noise and light levels. Good sleep quality is essential for overall quality of life. Poor sleep quality can have a number of detrimental impacts on one's physical as well as mental health. To improve sleep quality, it is important to establish a consistent sleep routine. There are many existing works on sleep quality prediction from wearable device data. Few of those analyzed sleep quality using the same algorithms used in this study. Several machine learning algorithms, however, have been proposed to reach great accuracy. Overfitting and insufficient data availability are common problems for these models. This research aims to increase the accuracy and performance of models for predicting sleep quality using wearable device data. To overcome these challenges, the objective of proposed work is to develop a sleep quality prediction system using a combination of feature selection techniques and machine learning models. The methodology is divided into three parts: data preprocessing, model building, and model evaluation. Three types of models were proposed in this study: single models, hybrid models, and an ensemble model for training and validation. The data acquired from a wearable IoT device was preprocessed by eliminating outliers and normalizing the data. The models were trained and evaluated based on accuracy, precision, recall, and F1-Score. The results show that the ensemble model was superior to all other models in terms of accuracy and F1-Score of 0.9897 and 0.9745 respectively. The hybrid models had lower performance metrics compared to the ensemble model, but still performed better than the individual models. This research provides insights into the potential of using wearable devices for sleep quality prediction and demonstrates the effectiveness of combining different models for improved accuracy and performance.