A case study on quality of sleep and health using Bayesian networks

Abstract

The objectives of this study are to investigate the associations of the socio-demographic characteristics, living habits, social and interpersonal factors and behaviour-related risk factors with both the quality of sleep and state of health. The tool used is Bayesian networks (BNs), a special case of probabilistic graphical models. The data utilised in this study is collected by employing the sample survey technique through the online network. A total of 1316 sets of data are collected with 20 variables of interest being studied. Our situation is whereby the BNs structure is unknown with full observability. A structural learning is conducted on the data to learn the correct network structure. There are several phases involved including implementation of the learning algorithms, integration of prior knowledge through the whitelist argument and arc setting to form directed acyclic graphs. The network scores computation is implemented to estimate the fitting of the resulting network of each structural learning algorithm in order to choose the best-fitted network. The arc strength or edge intensity is computed to estimate the marginal confidence on the presence of an arc. We found that the quality of sleep is dependent on certain factors in socio-demographic characteristics. The state of health is dependent on some factors in socio-demographic characteristics and living habits. We can also deduce that the quality of sleep has an impact on the state of health.