Related Experiment Video
Updated: Sep 6, 2025
04:04
Asthma Detection Research Based on Voice Signal Processing and Machine Learning
Published on: July 22, 2025
400
Enhancing PM2.5 Prediction Using NARX-Based Combined CNN and LSTM Hybrid Model
Ahmed Samy AbdElAziz Moursi1, Nawal El-Fishawy1, Soufiene Djahel2
1Computer Science and Engineering Department, Faculty of Electronic Engineering, Menoufia University, Menouf 32952, Egypt.
Sensors (Basel, Switzerland)
|June 24, 2022
Summary
This study introduces an advanced hybrid deep-learning model for hourly air pollution prediction, specifically targeting fine particulate matter (PM2.5). The novel CNN-LSTM/NARX approach significantly improves prediction accuracy, offering better environmental and health protection.
Area of Science:
- Environmental Science
- Computer Science
- Data Science
Background:
- Human expansion drives pollution, leading to air quality issues like climate change.
- Fine particulate matter (PM2.5) poses significant health risks, affecting respiratory and cardiovascular systems.
- Accurate short-term air pollution prediction is crucial for environmental and public health protection.
Purpose of the Study:
- To introduce an enhanced method for predicting particulate matter (PM2.5) concentrations within the next hour.
- To develop a hybrid deep-learning model integrating Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) with a Nonlinear Autoregression with Exogenous Input (NARX) model.
- To evaluate the model's performance against established metrics and state-of-the-art algorithms.
Main Methods:
- Development of a hybrid deep-learning architecture combining CNN and LSTM.
- Integration of the CNN-LSTM architecture with a Nonlinear Autoregression with Exogenous Input (NARX) model for enhanced time-series prediction.
- Evaluation of the proposed model using metrics such as Index of Agreement (IA) and Normalized Root Mean Square Error (NRMSE).
Main Results:
- The proposed CNN-LSTM/NARX hybrid model demonstrated superior performance in PM2.5 prediction.
- The model achieved the lowest Normalized Root Mean Square Error (NRMSE), indicating high accuracy.
- The Index of Agreement (IA) was highest for the proposed model, signifying excellent prediction capabilities.
- The hybrid model outperformed existing state-of-the-art deep-learning algorithms in PM2.5 forecasting.
Conclusions:
- The CNN-LSTM/NARX hybrid model offers a significant advancement in hourly air pollution prediction.
- This enhanced model provides a reliable tool for monitoring and mitigating the impacts of PM2.5.
- The findings highlight the potential of hybrid deep-learning approaches for environmental forecasting and public health preservation.