Scholarlink

The Air Quality Index (AQI) is a measure of the pollution levels in the air, usually developed by government agencies to convey the threat posed to the public by the pollution levels. In recent times, it has been observed that air pollution poses a major threat to public health during certain periods of the year. Since there is no universally accepted way of calculating AQI, this paper aims to predict daily AQI using a multi-pollutant index. Most of the research done on Indian data does not treat AQI as a time series but instead as a fully independent variable. To fill the gap in this area, an autoregressive approach to four classical machine learning algorithms, i.e., K-NN regressor, random forest, decision tree, and XGBoost was utilized. These models were trained on air quality data from metropolitan cities in India. The results have been evaluated in terms of the R2 score and absolute error and have been compared with state-of-the-art algorithms. An evolutionary approach to optimization was used to tune the hyperparameters of these models, utilizing a mixed continuous–discrete genetic algorithm, which drastically improved their performance. It was observed that AQI levels could be predicted at up to 97% accuracy using the genetically tuned XGBoost regressor.

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

An Evolutionary Approach for AQI Prediction Using Ensemble-Based Models

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Vikrant Arora,
  • Sudhanshu Marudgan,
  • Anoushka Ramankulath,
  • Punya Arora,
  • Preeti Gupta,
  • Sakshi Indolia

摘要

The Air Quality Index (AQI) is a measure of the pollution levels in the air, usually developed by government agencies to convey the threat posed to the public by the pollution levels. In recent times, it has been observed that air pollution poses a major threat to public health during certain periods of the year. Since there is no universally accepted way of calculating AQI, this paper aims to predict daily AQI using a multi-pollutant index. Most of the research done on Indian data does not treat AQI as a time series but instead as a fully independent variable. To fill the gap in this area, an autoregressive approach to four classical machine learning algorithms, i.e., K-NN regressor, random forest, decision tree, and XGBoost was utilized. These models were trained on air quality data from metropolitan cities in India. The results have been evaluated in terms of the R2 score and absolute error and have been compared with state-of-the-art algorithms. An evolutionary approach to optimization was used to tune the hyperparameters of these models, utilizing a mixed continuous–discrete genetic algorithm, which drastically improved their performance. It was observed that AQI levels could be predicted at up to 97% accuracy using the genetically tuned XGBoost regressor.