Scholarlink

<p>The prediction of air quality is important in preventing health risks as well as notifying early mitigation measures. To comply with this, the given paper introduces an enhanced temporal attention-based long short term memory (LSTM) model to forecast air quality. The proposed model uses mechanisms like feature engineering, model architecture adjustment and calibration technique to help increase model accuracy by focusing only on the most important time steps for PM2.5 (particulate matter) predictions. A real-time air quality data from Karnataka air pollution control board is used for testing the proposed model. The collected dataset undergoes feature scaling and processing of all variables followed by training and testing of model. The results show that the proposed model can be used as a valuable tool in early warning system empowering authorities to take proactive measures to minimize undesirable effect of air pollution on public health. The model attains the highest accuracy (97.53%) and lowest mean squared error (MSE) as 0.14, thus surpassing existing studies in terms of performance.</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

Enhanced temporal attention-based LSTM model for air quality forecasting

  • 期刊论文

  • 发表日期：

    2025年9月4日

查看全文
  • Mohamed Shakir,
  • U. Kumaran,
  • N. Rakesh

摘要

The prediction of air quality is important in preventing health risks as well as notifying early mitigation measures. To comply with this, the given paper introduces an enhanced temporal attention-based long short term memory (LSTM) model to forecast air quality. The proposed model uses mechanisms like feature engineering, model architecture adjustment and calibration technique to help increase model accuracy by focusing only on the most important time steps for PM2.5 (particulate matter) predictions. A real-time air quality data from Karnataka air pollution control board is used for testing the proposed model. The collected dataset undergoes feature scaling and processing of all variables followed by training and testing of model. The results show that the proposed model can be used as a valuable tool in early warning system empowering authorities to take proactive measures to minimize undesirable effect of air pollution on public health. The model attains the highest accuracy (97.53%) and lowest mean squared error (MSE) as 0.14, thus surpassing existing studies in terms of performance.