electricsheepafrica/african-malnutrition-dataset

这是一个用于非洲儿童严重急性营养不良（SAM）检测和共病筛查的合成数据集。数据集采用基于文献的概率建模方法，并结合赞比亚队列研究和DHS调查数据生成。SAM在非洲住院儿童中患病率为27%，且表现出毁灭性的共病模式，例如54.8%的HIV共感染和68.2%的结核共感染。数据集包含人体测量数据（WHZ评分、MUAC）、按年龄分层的患病率（6-24个月为峰值，占75.6%）以及显示SAM和非SAM儿童之间显著差异的血液学参数（平均血红蛋白分别为9.0 g/dL和10.6 g/dL）。四个数据集（共18,000个样本，3.0 MB）提供了用于生长监测、共病预测和干预目标配置。在这些数据上训练的模型预计将实现SAM检测的AUC-ROC >0.88和HIV/TB共感染筛查的AUC-ROC >0.75，为非洲营养不良计划中的社区健康工作者决策支持提供概念验证。

This is a synthetic dataset designed for the detection of severe acute malnutrition (SAM) and comorbidity screening in African children. The dataset is generated using probabilistic modeling informed by literature and cohort studies from Zambia and DHS surveys. It focuses on SAM, which affects a significant portion of hospitalized children in Africa, and its comorbidities, such as HIV and TB. The dataset includes anthropometric measurements, age-stratified prevalence, and hematological parameters. It is intended for growth monitoring, comorbidity prediction, and intervention targeting, with the expectation that models trained on this data will achieve high accuracy in SAM detection and comorbidity screening. The dataset is described in both English and provides information on its generation methodology, features, and limitations.