Unsupervised machine learning in air pollution epidemiology in South Africa

This dataset consist of different scripts and do files, used to achieve objectives to assess the applicability of machine learning in air pollution epidemiology in South Africa. The STATA do files were used to investigate the artificial intelligence (AI) survey distributed among postgraduate diploma students at the School of Health Systems and Public Health. R scripts were used for data imputation i.e., kalman, mice and mtsdi imputation, for the missing air pollution data and meteorological conditions. R scripts were also used for classification and regression trees to investigate joint effects of PM10, PM2.5, NO2, SO2 and O3 on respiratory and cardiovascular hospital admissions. Again presented are the R scripts for the unsupervised machine learning clustering methods i.e., k-means clustering, spectral clustering, dbscan clustering for joint effects for PM10, PM2.5, NO2, SO2 and O3 on respiratory and cardiovascular hospital admissions.

本数据集包含多种脚本和执行文件，旨在实现评估机器学习在南非空气污染流行病学适用性的目标。STATA执行文件被用于调查在健康系统与公共卫生学院研究生文凭学生中分发的人工智能（AI）调查。R脚本被用于数据补全，即卡尔曼滤波、多重插补（mice）和mtsdimputation，以补全缺失的空气污染数据和气象条件。R脚本亦被用于分类与回归树分析，以研究PM10、PM2.5、NO2、SO2和O3对呼吸和心血管住院的联合影响。此外，还提供了用于无监督机器学习聚类方法的R脚本，包括k-means聚类、谱聚类和DBSCAN聚类，以研究PM10、PM2.5、NO2、SO2和O3对呼吸和心血管住院的联合影响。