A Review on Causal Inference Models Combining Bayesian Networks with Deep Learning in Environmental Health Research

Abstract

This literature review explores the application of causal inference models, specifically Bayesian networks (BN), integrated with deep learning (DL) for feature selection, in assessing the association between air pollution exposure and disease prevalence among individuals living near industrial areas. Air pollution, especially in industrial zones, has been linked to a range of adverse health outcomes, including respiratory, cardiovascular, and chronic diseases. However, assessing these causal relationships remains a challenge due to the complexity of environmental and health data, as well as the presence of confounding factors. Traditional statistical methods often struggle to account for such complexity, which is where advanced models like Bayesian networks come into play. BNs, as probabilistic graphical models, offer a robust framework for modeling causal relationships, allowing for uncertainty and interaction between variables. Integrating deep learning techniques into Bayesian networks enhances feature selection, enabling the identification of critical factors influencing health outcomes while minimizing the impact of irrelevant or noisy variables. This paper reviews key studies that have employed these integrated models to investigate air pollution's health impact, focusing on the strengths, limitations, and potential of these methodologies. The review also highlights the challenges in modeling complex, real-world environmental health data and proposes directions for future research, including real-time data integration and enhanced computational methods. Ultimately, the combination of Bayesian networks and deep learning represents a promising approach for understanding and addressing the health impacts of air pollution in industrial areas.