International Journal of Engineering and Information Systems (IJEAIS)   Year: 2020 | Volume: 4 | Issue: 8 | Page No.: 88-101 An Automated LSTM-based Air Pollutant Concentration Estimation of Dhaka City, Bangladesh Rehnuma Karim and Taki Hasan Rafi Abstract: Dhaka is one of the mega-cities around the global. It has around 20 million population. Because of the outrageous population in less area, Dhaka is considered as one of the air polluted urban communities. So in this regard, air pollutant projection determining is a viable technique for ensuring general health by giving an early admonition against destructive air pollutants. It is a challenging task for researchers. However, existing automated strategies for air pollutant projection forecast is depend on term data acknowledgement. Which occurs forecast errors in many cases. In this paper, an efficient time-series artificial neural network model, Long-short term memory (LSTM) has been utilized for air pollutant concentration forecasting. Monthly PM2.5, PM10, SO2, NO2, CO2 and O3 concentration data collected at 3 air quality monitoring stations in Dhaka city from Jan-2013 to Jun-2020 were utilized to approve the viability of the proposed LSTM model. Additionally, the proposed model has been compared by autoregressive moving average (ARMA) and time-delay neural network (TDNN) model. To evaluate the viability of proposed model, mean absolute percentage error (MAPE) method has been utilized.