Scholarlink

<p>Rising temperatures due to climate change pose significant risks to the nutritional status of under-five children, particularly in Sub-Saharan Africa (SSA). This study investigates the influence of temperature increases on nutritional status (wasting, stunting, and underweight) in SSA. Based on Demographic and Health Survey (DHS) data for under-five children and global meteorological reanalysis data, we employed multiple supervised machine learning methods to predict the impact of temperature variability on nutritional status indicators, including stunting, underweight, and wasting, while controlling for socioeconomic variables such as household income and maternal education. Different metrics were used to evaluate the forecasting performance. In addition, multivariable logistic regression was employed to test for the causal-effect relationship. A total of 345,837 participants from 22 SSA countries were analyzed using data from 2005 to 2023. Among the algorithms tested, XG Boost achieved the highest accuracy for underweight prediction (Accuracy = 0.7832), Random Forest for stunting (Accuracy = 0.7023), and logistic regression for wasting (Accuracy = 0.6634). For different countries, accuracies ranging from 0.65 to 0.90, with highest in Uganda (decision tree, Accuracy = 0.9042 for stunting) and lowest in Burundi (XG Boost, Accuracy = 0.6426 for wasting). Causal-effect analysis revealed that each 1&#xa0;°C rise in average temperature increased the odds of stunting by approximately 1% (OR 1.01, 95% CI: 1.00–1.10), underweight by about 3% (OR 1.03, 95% CI: 1.01–1.06), and wasting by around 10% (OR 1.10, 95% CI: 1.08–1.12). Although the incremental increases per degree appear modest, such temperature-related risks may translate into substantial population-level impacts in climate-vulnerable settings. Higher household income and maternal education were associated with improved nutritional outcomes and attenuated the adverse effects of rising temperatures, indicating a protective socioeconomic effect. Supervised machine learning models can effectively leverage complex datasets to predict the impact of temperature variability on nutritional status, reinforcing the importance of integrated policies and climate-smart agricultural practices for safeguarding the health of under-five children in SSA.</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

Predicting the effects of temperature variability on nutritional status of children under five in Sub-Saharan Africa using machine learning

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年2月10日

查看全文
  • Jovine Bachwenkizi,
  • Cheng He,
  • Yixiang Zhu,
  • Alice Mugisha,
  • Boikhutso Tlou,
  • Candida Moshiro,
  • Henry Mwambi,
  • Isabel Madzorera,
  • Renjie Chen,
  • Haidong Kan,
  • Wafaie W. Fawzi

摘要

Rising temperatures due to climate change pose significant risks to the nutritional status of under-five children, particularly in Sub-Saharan Africa (SSA). This study investigates the influence of temperature increases on nutritional status (wasting, stunting, and underweight) in SSA. Based on Demographic and Health Survey (DHS) data for under-five children and global meteorological reanalysis data, we employed multiple supervised machine learning methods to predict the impact of temperature variability on nutritional status indicators, including stunting, underweight, and wasting, while controlling for socioeconomic variables such as household income and maternal education. Different metrics were used to evaluate the forecasting performance. In addition, multivariable logistic regression was employed to test for the causal-effect relationship. A total of 345,837 participants from 22 SSA countries were analyzed using data from 2005 to 2023. Among the algorithms tested, XG Boost achieved the highest accuracy for underweight prediction (Accuracy = 0.7832), Random Forest for stunting (Accuracy = 0.7023), and logistic regression for wasting (Accuracy = 0.6634). For different countries, accuracies ranging from 0.65 to 0.90, with highest in Uganda (decision tree, Accuracy = 0.9042 for stunting) and lowest in Burundi (XG Boost, Accuracy = 0.6426 for wasting). Causal-effect analysis revealed that each 1 °C rise in average temperature increased the odds of stunting by approximately 1% (OR 1.01, 95% CI: 1.00–1.10), underweight by about 3% (OR 1.03, 95% CI: 1.01–1.06), and wasting by around 10% (OR 1.10, 95% CI: 1.08–1.12). Although the incremental increases per degree appear modest, such temperature-related risks may translate into substantial population-level impacts in climate-vulnerable settings. Higher household income and maternal education were associated with improved nutritional outcomes and attenuated the adverse effects of rising temperatures, indicating a protective socioeconomic effect. Supervised machine learning models can effectively leverage complex datasets to predict the impact of temperature variability on nutritional status, reinforcing the importance of integrated policies and climate-smart agricultural practices for safeguarding the health of under-five children in SSA.