Toggle Main Menu Toggle Search

Open Access padlockePrints

Analyzing different dynamically modelled data structures and machine learning algorithms to predict PM2.5 concentration in China

Lookup NU author(s): Dr Jie ZhangORCiD

Downloads

Full text for this publication is not currently held within this repository. Alternative links are provided below where available.


Abstract

Harmful air pollutant such as PM2.5 is still a major concern in many countries. At high concentrations, it could lead to adverse health effect on human, escalating the risk of cardiovascular and respiratory diseases. In order to mitigate this issue, continuous air quality monitoring systems have been deployed to alert the general public of high PM2.5 level. However, such monitoring system requires substantial budget and resources to construct, thus may not be accessible in some regions especially developing countries. Therefore, it is important to develop a high performance PM2.5 prediction model that only employs easily attainable input parameters as a more cost-effective alternative. In this study, common meteorological data from five different cities in China were utilized for the PM2.5 prediction model. Dynamic model such as the nonlinear autoregressive network with exogenous inputs (NARX) with different input/output time lag were applied to transform training dataset into different data structures. Additionally, machine learning algorithms were analysed and evaluated to predict PM2.5, namely: multi-linear regression (MLR), and feed-forward artificial neural network (FANN). The results shows that FANN model with 10 hidden neurons using NARX-2 data structure is the best model combination with an R2 values of up to 0.973.


Publication metadata

Author(s): Djarum DH, Anuar NH, Ahmad Z, Zhang J

Editor(s): Yoshiyuki Yamashita, Manabu Kano

Publication type: Conference Proceedings (inc. Abstract)

Publication status: Published

Conference Name: Proceedings of the 14th International Symposium on Process Systems Engineering (PSE2021+)

Year of Conference: 2022

Pages: 1765-1770

Print publication date: 19/06/2022

Online publication date: 30/07/2022

Acceptance date: 22/02/2022

ISSN: 1570-7946

Publisher: Elsevier

URL: https://doi.org/10.1016/B978-0-323-85159-6.50294-3

DOI: 10.1016/B978-0-323-85159-6.50294-3

Series Title: Computer-Aided Chemical Engineering


Share