,
Associate Professor, Department of Artificial Intelligence and Data Science, Al-Ameen Engineering College (Autonomous) , Erode, Tamil Nadu , India
,
Professor, Department of Mathematics, Al-Ameen Engineering College (Autonomous) , Erode, Tamil Nadu , India
,
Assistant Professor, Department of Computer Science and Engineering, Al-Ameen Engineering College (Autonomous) , Erode, Tamil Nadu , India
,
Assistant Professor, Department of Computer Science and Engineering, Al-Ameen Engineering College (Autonomous) , Erode, Tamil Nadu , India
,
Assistant Professor, Department of Computer Science and Engineering, Al-Ameen Engineering College (Autonomous) , Erode, Tamil Nadu , India
,
Assistant Professor, Department of Computer Science and Engineering, Al-Ameen Engineering College (Autonomous) , Erode, Tamil Nadu , India
,
Assistant Professor, Department of Computer Science and Engineering, Al-Ameen Engineering College (Autonomous) , Erode, Tamil Nadu , India
Department of Computer Science and Engineering, Al-Ameen Engineering College (Autonomous) , Erode, Tamil Nadu , India
Hazardous air pollutants (HAPs) can be a critical risk to the sustainability of the environment and human health, which must be addressed by highly sophisticated predictive models to eliminate risks successfully. In this study, the researcher presents a hybrid deep learning model that combines Convolutional Neural Networks (CNNs) with the ability to extract the spatial features of the air quality with Long Short-Term Memory (LSTM) networks to learn the temporal relationships between air quality data. The study leverages the live Internet of Things (IoT) sensor data of urban and industrial areas in India where the researchers monitor the levels of , , , , temperature, and humidity. Principal Component Analysis (PCA) was used to select the best features that retain 95% of data variance; hence, the best model performance and lower redundancy were attained. The framework was strictly compared to baseline models in terms of such metrics as Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and latency. CNN-LSTM model showed great predictive performance, having an MAE of 3.2 μg/m3 and RMSE of 5.6 μg/m3, which were notably higher than those of Random Forest (MAE: 6.3 μg/m3) and XGBoost (MAE: 5.9 mu g/m3). Moreover, the Model registered the shortest prediction latency of 120 ms and a computational cost of 2.3 million FLOPs, which validated the Model to be real-time deployable. These findings demonstrate the possible role of deep learning in early warning systems, and further studies are focused on the enhancement of the approaches with reinforcement learning to manage pollution dynamically.
This is an open access article distributed under the Creative Commons Attribution Non-Commercial License (CC BY-NC) License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
0
The statements, opinions and data contained in the journal are solely those of the individual authors and contributors and not of the publisher and the editor(s). We stay neutral with regard to jurisdictional claims in published maps and institutional affiliations.