Hourly Air Quality Prediction in Dhaka City using Time Series Forecasting Techniques: Deep Learning Perspectives

Abstract

Air pollution is a concern worldwide, especially in densely populated cities in developing nations like Dhaka, Bangladesh. Accurately predicting air quality is crucial for health and environmental management. This research explores the effectiveness of learning models based on time series forecasting to predict air quality levels in Dhaka, with a particular focus on feature selection and the Air Quality Index (AQI). It also compares the performance and adaptability of models such as long-short-term memory (LSTM), convolutional neural network (CNN), recurrent neural network (RNN), and gated recurrent unit (GRU) using various feature configurations. The study highlights that LSTM performs well in scenarios while CNN, RNN and GRU consistently exhibit strong predictive capabilities. Underscore their suitability for air quality forecasting particularly when considering the prediction of AQI. However, the study acknowledges the limitations of relying on data sources. Emphasizes the need for further research into hybrid models and how local factors impact air quality. The findings provide insights for practitioners and decision-makers aiming to tackle air pollution challenges in urban areas by emphasizing feature selection and AQIs role in enhancing predictive accuracy.