Prediction of particulate matters at urban intersection by using multilayer perceptron model based on principal components

Research output: Journal Publications and Reviews (RGC: 21, 22, 62)21_Publication in refereed journalpeer-review

16 Scopus Citations
View graph of relations



Original languageEnglish
Pages (from-to)2107-2114
Journal / PublicationStochastic Environmental Research and Risk Assessment
Issue number8
Publication statusPublished - 25 Nov 2014


The time series of particulate matter at urban intersection consists of complex linear and non-linear patterns and are difficult to predict. Multilayer perceptron (MLP) model has been applied to air quality prediction in urban areas, but it has limited accuracy owing to the co-linearity between the input variables. To overcome it, a novel hybrid model combining MLP model and principal component analysis (PCA) is proposed to improve the prediction accuracy. The PCA was applied before the MLP model was implemented to generate principal components as input variables, rather than using the original data, to reduce the complexity and eliminate data co-linearity. The proposed model is examined with the measured data of particulate matter (PM) concentrations in spring and winter respectively. Experimental results indicated that the hybrid model can be an effective tool to improve the particulate matter prediction accuracy. Additionally, the proposed model is found to perform better for analyzing PM1 levels than PM10. The model is also verified to produce better predictions in winter than that in spring. The outputs of these findings demonstrate the potential of the proposed model to be applied to predict the trends of air pollution in similar meso- to mega- cities.

Research Area(s)

  • Multilayer perceptron, Particulate matter, Principal components, Traffic emissions

Citation Format(s)