Temperature Prediction Based on EAL Deep Learning Model

Temperature prediction has certain reference significance for future warming and related extreme weather. The existing traditional models cannot capture the features of temperature time series. In order to solve the problem of low accuracy of temperature time series prediction by traditional models, this paper proposes an EAL-based deep learning model to capture its features. In this paper, firstly, ARIMA and LSTM models were constructed and the temperature series were analyzed, and it was found that the prediction results were poor. Therefore, in this paper, with the help of EEMD decomposition, the fused temperature prediction model EAL is proposed, and finally a good fitting and prediction result is obtained. The three models of LSTM, ARIMA and EAL were compared, where the RMSE was 0.2912, 0.1064 and 0.0546, MAE was 0.1818, 0.0863 and 0.0277, and MAPE was 0.0131, 0.0061 and 0.002, respectively. Therefore, it was found that the EAL model in all indicators was outperformed other models and had better applicability for this field of temperature prediction.