Sleep Quality Prediction and Estimation by using Deep Belief Networks-Bayesian Optimization

Sufficient sleep is important for individuals who are suffering with chronic diseases, as sleep scarcity severely impacts emotional, physical and mental well-being, worsening health complications. The existing convolutional network for sleep quality prediction and estimation has computational complexity and over fitting by which the model didn’t effectively predict the quality of sleep. Therefore, this research proposes Deep Belief Networks-Bayesian Optimization (DBN-BO) for predicting and estimating the quality of sleep. Initially, Sleep study data is used as input and this data contains outliers which are preprocessed by using local outlier factor and effectively reduced impact of noise while detecting the outliers. Then, the sleep quality is predicted by using DBN which learns the complex features from raw data. After that, BO is used for estimating the quality of sleep and explores the various factors related to sleep. This proposed DBN-BO effectively reduced the computational complexity and over fitting by training one layer at a time, proposed DBN-BO achieved better results in terms of accuracy $\mathbf{9 4 . 4 5 \%}$, sensitivity $\mathbf{9 7 . 0 9 \%}$, and specificity of $\mathbf{9 3 . 4 5 \%}$ and F1-Score of $91.12 \%$ when compared with existing convolutional layers.