Air Quality Predictions in Urban Areas Using Hybrid ARIMA and Metaheuristic LSTM

Abstract

Due to the development of transportation, population growth and industrial activities, air quality has become a major issue in urban areas. Poor air quality leads to rising health issues in the human’s life in many ways especially respiratory infections, heart disease, asthma, stroke and lung cancer. The contaminated air comprises harmful ingredients such as sulfur dioxide (SO2), nitrogen dioxide (NO2), and particulate matter of PM10, PM2.5, and an Air Quality Index (AQI). These pollutant ingredients are very harmful to human’s health and also leads to death. So, it is necessary to develop a prediction model for air quality as regular on the basis of monthly or seasonaly. In this work, a new hybrid model for air quality prediction (AQP) is developed by using reed deer metaheuristic optimized Long Short Term Memory (LSTM) Deep Learning network. To overcome the drawback of the existing autoregressive integrated moving average model (ARIMA) model, the residual errors are processed by using an optimized LSTM network. The red deer optimization (RDO) is a new type of metaheuristic method which is motivated by the mating behaviour of Red Deer. The proposed model is better in terms of all prediction performance parameters when compared with other models.