Enhanced air quality index prediction using a hybrid convolutional network

Pei-Chun Lin, Pei-Chun Lin and Arbaiy, Nureize and Yu, Chen-Yu and Mohd Salikon, Mohd Zaki (2024) Enhanced air quality index prediction using a hybrid convolutional network. In: Recent Advances on Soft Computing and Data Mining.

[img] Text
P17174_dfe5174c6d5badc8744a78af722c8558.pdf
Download (8MB)

Abstract

Accurate air quality forecasting is critical for decreasing pollution and protecting public health. A hybrid model combining the Temporal Convolution Network (TCN) and the Graph Convolution Network (GCN) has been developed to predict air pollution with high accuracy and minimise the associated health risks. Because air quality data has two crucial components: temporal trends and spatial linkages, the combination of TCN and GCN is required. The GCN model learns the complicated architecture of each observatory, whereas the TCN model uses past data to detect deviations. The Graph Temporal Convolution Network (GTCN) model was evaluated using six important variables: station names, Air Quality Index (AQI), data timestamps, longitude, and latitude. Our GTCN outperformed other researchers’ models on real-world data between February and July 2021. The results demonstrated the lowest Mean Absolute Error (MAE) of approximately 4.78 and the lowest Root Mean Square Error (RMSE) of approximately 6.67. Through precise air quality forecasting, people can pre-know how to protect themselves and prepare outdoor dresses well to reduce exposure to air pollution and related health hazards

Item Type: Conference or Workshop Item (Paper)
Uncontrolled Keywords: Air Quality Index, Graph Convolution Network, Temporal Convolution Network, Uncertainty, Prediction Model
Subjects: T Technology > TD Environmental technology. Sanitary engineering
Divisions: Faculty of Computer Science and Information Technology > FSKTM
Depositing User: Mrs. Sabarina Che Mat
Date Deposited: 09 Jan 2025 08:15
Last Modified: 09 Jan 2025 08:16
URI: http://eprints.uthm.edu.my/id/eprint/11942

Actions (login required)

View Item View Item