Air pollutant prediction based on ARIMA-WOA-LSTM model

Abstract

The problem of air pollution has always plagued people's lives, and the management of air pollution cannot be achieved without the prediction and assessment of the concentration of various pollutants. In this paper, we propose a method to accurately predict air pollutants with the aim of ensuring the efficiency of air pollution management. The proposed ARIMA-WOA-LSTM model uses ARIMA to extract the linear part of air pollution data and output the nonlinear part, while the WOA-LSTM model is used to predict the nonlinear part, where the whale algorithm is used to find the perfect hyperparameters for the LSTM, and the objectives of the search include the number of neurons, model learning rate and batch length. To prove the excellence of the model developed in the article ARIMA-WOA-LSTM is compared with ARIMA-LSTM, CEEMDAN-WOA-LSTM, WOA-LSTM, ARIMA, and LSTM. The results show that ARIMA-WOA-LSTM performs better than other models in three aspects: pollutant prediction accuracy, model prediction accuracy, and prediction stability; the combined model also performs much better than the single model in the three aspects; The whale algorithm is excellent for the search of the five hyperparameters in the LSTM, which is important for the error reduction of the model. ARIMA-WOA-LSTM model has high reference for air pollution management.