A Bayesian Network Approach for Discovering Variables Affecting Youth Depression

Abstract

Bayesian Networks have been used for data mining in many domains, but they have been rarely adopted in educational domain. In this paper, we model a Bayesian Network to discover which variables are in charge of youth depression and how strong the variables influence. For this study, Korean Children and Youth Panel Survey data are used and Markov Blanket is adopted to learn the Bayesian Network and to choose the relevant variables. In the results, “life satisfaction”, “social withdrawal”, “mobile phone dependency”, “attention”, “caregiver abuse”, and “aggressiveness” are extracted as the relevant variables to youth depression, therefore caregivers should pay attention to these variables of youths to reduce their depression. This paper shows Bayesian Networks are quite effective in finding the causal variables and their effects in educational domain.