Scholarlink

<p>Globally, industrialization has driven a socio-economic transformation that catalyzed the development of the modern technological era. However, despite its benefits for quality of life (QoL) and social well-being, its collateral effects have been severe; excessive emission of greenhouse gases intensifies climate instability, exacerbates environmental pollution, and increases public health risks. In fact, the WHO reports 7 million annual premature deaths, 80% linked to poor air quality (AQ). In Central America, the situation is critical due to socioenvironmental vulnerability, particularly in regions such as the Atmospheric Basin of Tula (ABT) in Mexico, an industrial area declared a sanitary emergency due to its lethality and epidemiological risk. The current scenario reveals three structural trends of growing uncertainty, two adverse; (1) environmental and (2) socio-health, and a third aimed at addressing them; the technology through deep learning (DL) neural models with evolutionary preventive potential. This study analyzed environmental conditions, health risks, and QoL in ABT, based on seven alarming findings, seven DL models (three recurrent, two of classification, and two of regression) were developed applying the CRISP-DM methodology. CO<sub>2</sub> achieved R<sup>2</sup> = 0.82, Explained Variance (EV)= 0.80, and RMSE = 5.07; O<sub>3</sub> + NO<sub>2</sub> obtained R<sup>2</sup> = 0.88, EV = 0.88, and RMSE = 16.44; PM<sub>10</sub> reached R<sup>2</sup> = 0.98, EV = 0.99, and RMSE = 20.89; blood pressure (BP) showed R<sup>2</sup> = 0.99, EV = 0.98, and RMSE = 0.00012; for chronic disease (CD) risk and AQ achieved accuracy rates of 98%, 98%, sensitivity of 97%, 93%, and F1-score of 98%, 95%, respectively; QoL attained R<sup>2</sup> = 0.99, EV = 0.99, and RMSE = 0.000974, scores of 0.00115 physical health, 0.002469 psychological, 0.000981 social relationships, and 0.000974 environment. DL proves to be a structural trend capable of representing and predicting dynamic health-ecological phenomena, this is an unprecedented preventive tool for confronting complex future scenarios, reducing risks, and supporting decision-making toward plausible outcomes.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Air pollution and deep learning in prevention, new challenges and new solutions

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年12月13日

查看全文
  • Silvia Soledad Moreno-Gutiérrez,
  • Héctor Hugo Siliceo-Cantero,
  • Ismael Domínguez-Jiménez

摘要

Globally, industrialization has driven a socio-economic transformation that catalyzed the development of the modern technological era. However, despite its benefits for quality of life (QoL) and social well-being, its collateral effects have been severe; excessive emission of greenhouse gases intensifies climate instability, exacerbates environmental pollution, and increases public health risks. In fact, the WHO reports 7 million annual premature deaths, 80% linked to poor air quality (AQ). In Central America, the situation is critical due to socioenvironmental vulnerability, particularly in regions such as the Atmospheric Basin of Tula (ABT) in Mexico, an industrial area declared a sanitary emergency due to its lethality and epidemiological risk. The current scenario reveals three structural trends of growing uncertainty, two adverse; (1) environmental and (2) socio-health, and a third aimed at addressing them; the technology through deep learning (DL) neural models with evolutionary preventive potential. This study analyzed environmental conditions, health risks, and QoL in ABT, based on seven alarming findings, seven DL models (three recurrent, two of classification, and two of regression) were developed applying the CRISP-DM methodology. CO2 achieved R2 = 0.82, Explained Variance (EV)= 0.80, and RMSE = 5.07; O3 + NO2 obtained R2 = 0.88, EV = 0.88, and RMSE = 16.44; PM10 reached R2 = 0.98, EV = 0.99, and RMSE = 20.89; blood pressure (BP) showed R2 = 0.99, EV = 0.98, and RMSE = 0.00012; for chronic disease (CD) risk and AQ achieved accuracy rates of 98%, 98%, sensitivity of 97%, 93%, and F1-score of 98%, 95%, respectively; QoL attained R2 = 0.99, EV = 0.99, and RMSE = 0.000974, scores of 0.00115 physical health, 0.002469 psychological, 0.000981 social relationships, and 0.000974 environment. DL proves to be a structural trend capable of representing and predicting dynamic health-ecological phenomena, this is an unprecedented preventive tool for confronting complex future scenarios, reducing risks, and supporting decision-making toward plausible outcomes.