Journal of Dali University ›› 2023, Vol. 8 ›› Issue (6): 44-51.

Previous Articles     Next Articles

Dali Climate Prediction Based on Long Short-Term Memory Network Model

Yuan Jincheng， Li Zhongmu*   

  1. Institute for Astronomy， Dali University， Dali， Yunnan 671003， China
  • Received:2022-08-29 Revised:2022-12-03 Online:2023-06-15 Published:2023-06-27

PDF (PC)

40

Abstract:

Based on the Keras framework， a long short-term memory network （LSTM） model is constructed for climate prediction in Dali. The experimental results showed that the model could accurately predict the temperature， precipitation， humidity， and sunshine hours from June 2016 to May 2021， with root mean square error（RMSE） of 1.544 ℃， 2.720 mm， 6.521% rh， and 1.990 h， and mean absolute percentage error（MAPE） of0.087 ℃， 4.025 mm， 0.085% rh， and 0.462 h. The model performs well in predicting temperature and humidity， but the prediction error is relatively large for precipitation and sunshine hours due to their large fluctuations and extreme values. The RMSE and MAPE obtained by the experiment are lower than those obtained by traditional models such as recurrent neural network  （RNN）， support vector machine （SVM）， and K-nearest neighbor （KNN）. Therefore， the LSTM model can produce higher prediction accuracy for climate prediction in Dali.

Key words: ">machine learningfont-family:宋体, ">, ">LSTMfont-family:宋体, ">, ">climate predictionfont-family:宋体, ">, ">climate change

CLC Number: