Protective consumption behavior under smog: using a data-driven dynamic Bayesian network.

Abstract

In the midst of the deteriorating air pollution and collective stress, people pay close attention to risk mitigation measures such as keeping indoor and purchasing anti-smog products. Through impact evaluations, factors regarding health protective behavior can be identified. However, limited research is available regarding probabilistic interdependencies between the factors and protective behavior and largely relies on subjective diagnosis. These concerns have led us to adopt a data-driven static Bayesian network (BN) and Dynamic BN model to help explore multidimensional factors that may influence the public's health protective behavior of buying anti-smog air purifiers and examine the dependencies among network nodes. Using the city-level aggregate data from an online shopping platform, the results shed new light on relationships existing among 11 factors and protective behavior of buying air purifiers. Furthermore, taking into account the dynamic nature of protective behavior, we add time-related factors on the basis of static BN to construct the dynamic BN model. Results indicate that PM2.5 concentration and product price are the two leading factors affecting the consumption behavior for air purifiers. Additionally, media-related factors play an important role in the consumption behavior. This study contributes to the fields of impact evaluation of protective consumption behavior analysis and links environment risk with public consumption by identifying key factors.