Use of machine learning algorithms to determine the relationship between air pollution and cognitive impairment in Taiwan

Cheng-Hong Yang,Chih-Hsien Wu,Kuei-Hau Luo,Huang-Chih Chang,Sz-Chiao Wu,Hung-Yi Chuang

DOI: https://doi.org/10.1016/j.ecoenv.2024.116885

2024-10-01

Abstract:Air pollution has become a major global threat to human health. Urbanization and industrialization over the past few decades have increased the air pollution. Plausible connections have been made between air pollutants and dementia. This study used machine learning algorithms (k-nearest neighbors, random forest, gradient-boosted decision trees, eXtreme gradient boosting, and CatBoost) to investigate the association between cognitive impairment and air pollution. Data from the Taiwan Biobank and 75 air-pollution-monitoring stations in Taiwan were analyzed to determine individual levels of exposure to air pollutants. The pollutants examined were particulate matter with a diameter of ≤ 2.5 μm (PM2.5), nitrogen dioxide, nitric oxide, carbon monoxide, and ozone. The results revealed that the most strongly correlated with cognitive impairment were ozone, PM2.5, and carbon monoxide levels with adjustment of educational level, age, and household income. The model based on these factors achieved accuracy as high as 0.97 for detecting cognitive impairment, indicating a positive association between air pollutions and cognitive impairment.