Scholarlink

<p>Pollution in whole is a prominent health affecting factor; this paper focuses on air pollution in particular in pollution. Air pollution is posing a serious health threat in metropolitan cities such as Delhi NCR. Prediction of accurate Air Quality Index (AQI) is necessary for timely precautions and awareness among locals. This study focuses on a stacked ensemble AQI prediction framework using two publicly available dataset, Delhi Weather and AQI dataset for regression task (training and testing using 5-fold cross validation) and Delhi Pollution AQI Dataset for independent validation and classification task of mapping AQI values to health risk categories defined by Central Pollution Control Board. The proposed framework gives strong predictive performance with average MAE of 12.7, RSME of 17.9 and R<sup>2</sup> of 0.95 for primary dataset as well as maintain robust performance with MAE of 14.2, RSME of 19.6 and R<sup>2</sup> of 0.93 reflecting generalization capability. This study also mapped predicted AQI with health risk categories which allows system to classify pollution severity levels. The classification task achieved accuracy of 81.3% representing ability to effectively identify pollution risk levels. Furthermore, a comparison analysis was also done with several existing model using common dataset, Delhi Pollution AQI dataset. The proposed stacked ensemble framework outperformed the existing model by giving lower prediction errors and higher value of R<sup>2</sup>.</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

A life-cycle oriented stacked-ensemble framework for air quality prediction and health risk analysis

  • 期刊论文

  • 发表日期：

    2026年8月4日

查看全文
  • Sonal Shukla,
  • Sambit Satpathy,
  • Dhirendra Kumar Shukla,
  • Mamta Narwaria,
  • Mohit Chowdhary

摘要

Pollution in whole is a prominent health affecting factor; this paper focuses on air pollution in particular in pollution. Air pollution is posing a serious health threat in metropolitan cities such as Delhi NCR. Prediction of accurate Air Quality Index (AQI) is necessary for timely precautions and awareness among locals. This study focuses on a stacked ensemble AQI prediction framework using two publicly available dataset, Delhi Weather and AQI dataset for regression task (training and testing using 5-fold cross validation) and Delhi Pollution AQI Dataset for independent validation and classification task of mapping AQI values to health risk categories defined by Central Pollution Control Board. The proposed framework gives strong predictive performance with average MAE of 12.7, RSME of 17.9 and R2 of 0.95 for primary dataset as well as maintain robust performance with MAE of 14.2, RSME of 19.6 and R2 of 0.93 reflecting generalization capability. This study also mapped predicted AQI with health risk categories which allows system to classify pollution severity levels. The classification task achieved accuracy of 81.3% representing ability to effectively identify pollution risk levels. Furthermore, a comparison analysis was also done with several existing model using common dataset, Delhi Pollution AQI dataset. The proposed stacked ensemble framework outperformed the existing model by giving lower prediction errors and higher value of R2.