Scholarlink

The article demonstrates application of nonlinear autoregressive neural network for multi-step ahead hourly prediction of concentration of Particulate Matter (PM2.5) at Anand Vihar location in New Delhi, India. The hourly PM2.5 concentration data of the area has been collected for a duration of 10 days and used for training of proposed machine learning model. The autoregressive neural network based machine learning model has been trained using three different training algorithms namely Levenberg–Marquardt (LM), Bayesian Regularization (BR) and Scaled Conjugate Gradient (SCG). The prediction output of these algorithms have been analysed for a maximum hourly prediction duration of 5 h in terms of root mean squared error (RMSE), mean absolute error (MAE) and mean absolute percentage error (MAPE). The results clearly indicate that BR-algorithm outperforms other machine learning algorithms considered in the study.

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

Multistep-Ahead Prediction of PM2.5 Using Nonlinear Autoregressive Machine Learning Model

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Ashwani Kharola,
  • Kapil Kumar Sharma,
  • Amir Shaikh,
  • Ravi Kanojia,
  • Shristi Kharola,
  • Deepak Juyal,
  • Vishwjeet Choudhary,
  • Tarun Kumar Dhiman,
  • Kiran Sharma,
  • Vinayak Sharma

摘要

The article demonstrates application of nonlinear autoregressive neural network for multi-step ahead hourly prediction of concentration of Particulate Matter (PM2.5) at Anand Vihar location in New Delhi, India. The hourly PM2.5 concentration data of the area has been collected for a duration of 10 days and used for training of proposed machine learning model. The autoregressive neural network based machine learning model has been trained using three different training algorithms namely Levenberg–Marquardt (LM), Bayesian Regularization (BR) and Scaled Conjugate Gradient (SCG). The prediction output of these algorithms have been analysed for a maximum hourly prediction duration of 5 h in terms of root mean squared error (RMSE), mean absolute error (MAE) and mean absolute percentage error (MAPE). The results clearly indicate that BR-algorithm outperforms other machine learning algorithms considered in the study.