Toggle Main Menu Toggle Search

Open Access padlockePrints

Autoencoder artificial neural network model for air pollution index prediction

Lookup NU author(s): Dr Jie ZhangORCiD

Downloads


Abstract

Air pollution is associated with the release of air pollutants in terms of gases, particulates and biological molecules into the atmosphere which is detrimental to the health. It is mainly caused by the industrialisation, rapid urbanisation and population growth. It is one of the fundamental problems faced in worldwide scale. Particularly, power plant energy production, industrial processes, fuel burning vehicles and residential heating are some of the primary causes of the problem. Thus, the air quality monitoring and forecasting tools are definitely essential so that precautionary measures can be taken by minimising the potential negative effects of predicted pollution peaks onto the surrounding ecosystem. In this paper, shallow sparse autoencoder and deep sparse autoencoder artificial neural network models are proposed to enhance the modelling prediction and forecasting of Air Pollution Index (API) in Perak Darul Ridzuan State in Malaysia as a case study. The result shows that deep sparse autoencoder can further improve the API model prediction performance with MSE and R2 values of 0.1474 and 0.8331, respectively as compared to shallow sparse autoencoder with MSE and R2 values of 0.1515 and 0.8300, respectively. The prediction performances obtained using shallow and deep sparse autoencoders are further compared with other models proposed in previous research in predicting API, which are feedforward artificial neural network (FANN) and principal component analysis (PCA) models in order to validate the developed autoencoder models. It clearly shows that both shallow and deep sparse autoencoders do improve the API prediction performance with deep sparse autoencoder being selected as the final architectural structure.


Publication metadata

Author(s): Basir NI, Tan KK, Djarum DH, Ahmad Z, Vo DVN, Zhang J

Publication type: Article

Publication status: Published

Journal: IIUM Engineering Journal

Year: 2025

Volume: 26

Issue: 1

Pages: 1-21

Print publication date: 10/01/2025

Online publication date: 10/01/2025

Acceptance date: 18/09/2024

Date deposited: 22/09/2024

ISSN (print): 1511-788X

ISSN (electronic): 2289-7860

Publisher: International Islamic University Malaysia

URL: https://doi.org/10.31436/iiumej.v26i1.2818

DOI: 10.31436/iiumej.v26i1.2818


Altmetrics

Altmetrics provided by Altmetric


Funding

Funder referenceFunder name
FRGS/1/2022/TK05/USM/01/5
Kementerian Pendidikan Malaysia (KPM)

Share