Dynamical prediction of two meteorological factors using the deep neural network and the long short term memory (1).

摘要: It is important to calculate and analyze temperature humidity prediction accuracies among quantitative meteorological forecasting. This study manipulates the extant neural network methods foster predictive accuracy. To achieve such tasks, we explore accuracy performance in networks using two combined factors (temperature humidity). Simulated studies are performed by applying artificial (ANN), deep (DNN), extreme learning machine (ELM), long short-term memory (LSTM), with peephole connections (LSTM-PC) methods, accurate value compared that obtained from each other methods. Data extracted low frequency time-series of ten metropolitan cities South Korea March 2014 February 2020 validate our observations. test robustness error LSTM found outperform four Particularly, as testing results, summer Tongyeong has a root mean squared (RMSE) 0.866 lower than while absolute percentage (MAPE) for 5.525 Mokpo, significantly better cities.