Weather prediction, ARIMA model, SARIMA model, SARIMAX model
Abstract
Weather predicting is important nowadays as it can not only significantly diminish the uncertainty regarding the future, but also provide valuable information for people to make decisions in advance in different areas. ARIMA, SARIMA and SARIMAX models are typical predicting models, which are favored and used by scholars from various nations. This essay aims to compare the predicting performance of them through data processing, parameters selecting, performance measuring and so on. The conclusion is as follows: the seasonal factors incorporated by SARIMA model can greatly improve the accuracy of prediction with smaller errors compared with ARIMA model. Besides, when considering the exogenous variables like longitude and latitude, the predicting performance of models can be enhanced as well. These findings can provide good suggestions for further research. Therefore, research in the future is recommended to exploit the full potential of SARIMAX model in other areas with more exogenous variables and also can attempt to find the best method for selecting exogenous variables.
