Year: 2016
Type: conference-paper
Abstract: Bayesian Networks are networks of interconnected variables used to explain causal relationships with conditional probability. Latent variables or hidden variables are variables that cannot be directly measured, like depression or physical activity. They can be used inside of a Bayesian Network. This research looks at latent variables as a weighted sum of observed variables. We use these modeled latent variables as continuous variables in a Bayesian Network. As an example, we look at a Bayesian Network of the causation of diabetes using data from the National Health and Nutrition Examination Survey (NHANES) that is publicly available from the CDC and contains several health-related variables. In this example, we model physical inactivity, as a weighted sum of variables in the data. We found that physical inactivity can be modeled as a linear combination of the total number of hours watching TV and the number of hours spent doing vigorous/moderate physical activity. These variables were inversely correlated, meaning we contrasted amount of activity with time spent watching TV.
Author: Charles A. Cain
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
Funder: University of Minnesota
OpenAlex ID: W2346776522