Prediction of pm10 concentration using multiple linear regression and support vector machine
Loading...
Date
2018-06
Authors
Masezatti Zailan
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Particulate matter with an aerodynamic diameter less than 10µm (PM10) is one of
the most air pollutants that can give negative effect on human health and environment.
The purpose of this research is to predict the particulate matter concentration for the next
day (PM10D1) by using Multiple Linear Regression (MLR) and Support Vector Machine
(SVM) models. The meteorological and gaseous parameters that are used in this study
are particulate matter for today (PM10D0), wind speed (WS), temperature (TEMP),
relative humidity (RH), sulphur dioxide (SO2), nitrogen dioxide (NO2), ozone (O3) and
carbon monoxide (CO). The daily mean data that are used in this study are divided into
training data (70%) and validation data (30%) and are used from 2013 until 2015. Four
monitoring stations were selected in this study to predict the PM10 concentration for the
next day (PM10D1) which are Jerantut which act as background station, Nilai (industrial
area), Seberang Jaya (sub-urban area) and Shah Alam (urban area). The results of overall
data that are obtained from this study has shown that Nilai monitoring stations
contributed the highest mean value of PM10 concentration compared to the other
monitoring stations. This indicated that Nilai is a more polluted area as it is known as a
highly industrialised area. The results shows that Multiple Linear Regression (MLR) is
the best model in predicting PM10 concentration for the next day compared to Support
Vector Machine (SVM) model.