Air Quality Forecasting using Temporal Convolutional Network (TCN) Deep Learning Method

Abstract

This study aims to build an air quality model to predict pollutant concentrations in Malaysia. The method chosen in this study is one of the deep learning techniques which is the temporal convolution network (TCN). The data set used is particulate matter with diameter of 10 micrometers or less (PM10) time series which is obtained from the Department of Environment Malaysia from 5th July 2017 to 31st January 2019. Data from five air quality monitoring stations in Peninsular Malaysia were selected for this study. The long-short term memory network (LSTM) is also used in this study for the purpose of accuracy comparison between the two models. Overall, the forecast values from both models are approximately close to the original data. However, the TCN model is better in terms of the forecast accuracy. This study shows that TCN is a suitable technique that can be used for forecasting air quality time series data in Peninsular Malaysia.