Sustainable optimized LSTM-based intelligent system for air quality prediction in Chennai

Nowadays, air quality prediction is the most essential process taken by an Indian government. Due to poor quality of air, unhealthy lifestyle and premature deaths of humans have arisen in India, especially in Delhi. Not only has a human’s health, but the air pollution also made a huge impact on several areas like economy, agriculture and road accidents, etc. In recent times, deep learning (DL) technologies are influenced every application rapidly even in air pollution prediction. In this work, the novel optimised DL algorithms are proposed for the efficient prediction of air quality particularly focussing on Chennai, Tamil Nadu. To provide higher accuracy in air quality prediction, the novel optimised DL algorithms are proposed which is combined several models like ARIMA and CNN-LSTM and Tuna Optimization Algorithm, respectively. Initially, CNN and LSTM are combined to provide hybrid architecture. Next, the metaheuristics-based tuna swarm optimization model is applied for fine-tuning the hyperparameters of the CNN-LSTM model which is known as the Tuna Optimised CNN-LSTM (TOCL) method. Finally, the novel TOCL is applied to the residuals of the ARIMA model to form an ARIMA- TOCL (ARTOCL) model. As a result, the novel ARTOCL is learned and performed with an optimal air quality prediction. The metrics of the Hybrid ARTOCL model are evaluated as a better mean absolute error (MAE), root mean squared error (RMSE), R2 score and the normalized RMSE (nRMSE) with higher accuracy than the previous models. The results show that the proposed prediction model has 22.6% R2 improvement, 14.6% MAE reductions, 22% RMSE reductions and 16.45% nRMSE reductions than the existing models.