Research article

PREDICTION OF ATMOSPHERIC TEMPERATURE IN BEIJING BASED ON ARTIFICIAL NEURAL NETWORK

Qingchun Guo1,2,3*, Zhenfang He1,2,4*, Zhaosheng Wang5*

Online First: December 30, 2022


Climate change affects the growth and development of plants and has an important impact on forests. We use artificial neural network (ANN) to predict the change of monthly atmospheric temperature in Beijing City. We divide the atmospheric temperature data from 1951 to 2021 into training data sets (1951-2007) and prediction data sets (2008-2021), and use the 20 month data as model input to predict the atmospheric temperature in the next month. Thirteen various training functions were trained and compared. The ANN model was tested by correlation coefficient (R), root mean square error (RMSE), and mean absolute error (MAE) and good results were obtained. Bayesian Regularization (trainbr) is the best performing algorithms compared to other algorithms with R value of 0.9977 and the lowest error values for RMSE (0.7557), MAE (0.6173). This method has good generalization ability and can be used to forecast atmospheric temperature in other regions.

Keywords

atmospheric temperature, artificial neural network, prediction, training algorithm, forest.