Antarctic Temperature Prediction Based on the Long Short Term Memory Network

In the context of global warming, Antarctic temperature prediction plays a significant role in global warming and the environment, which deserves more attention due to the rapid melting of the ice sheet in Antarctica as temperatures rise. In this study, the temperature tables of 18 stations in Antarctica are collected firstly on the Kaggle website, and secondly, the features of each data are staggered using a double layer cycle to process the missing values and extract the complete Antarctic temperature values. And the data are processed as training and samples using LSTM, ARIMA, SARIMA, linear regression, and decision regression tree modules. Through the experimental results, the Mean Absolute Error (MAE) of Antarctic temperature prediction by LSTM is 1.91 and the Mean Squared Error (MSE) is 4.71. Compared with the other four models, the model prediction effect of LSTM is better and more suitable for predicting Antarctic temperature, and the last part concludes that the advantages and disadvantages of LSTM, ARIMA, SARIMA, linear regression, and decision regression tree modules are analyzed. The significance of this study is LSTM prediction of Antarctic temperature prediction can help people to understand and recognize the global warming trend intuitively and the prediction models in this paper are meaningful in predicting temperatures more quickly than traditional prediction models. In addition, the data from this analysis can be applied to various fields and prompt people to pay attention to the change in the Antarctic temperature.