Time Series Temperature Prediction Model Based on Big Data

Global warming relates to human life and death, and an important indicator of climate is temperature. This paper studied the annual average temperature change trend of Chinese land based on time series models. The data period is 1838-2021 (a total of 184 years). After comparing the models, an ARIMA model was selected for modeling and prediction. For choosing the autoregressive coefficient p and the moving average coefficient q, the Akaike information criterion (AIC) and the Bayesian information criterion (BIC) were applied to simulate various situations, and then the comprehensive minimum value was determined. The results show that the model is the best for ARIMA (2,1,5). Finally, the model was implemented to predict the temperature from the year 2022 to 2030. To test the accuracy of the model, the root mean square error (RMSE=0.26857) and the mean absolute error(MAE=0.21371) between the predicted values and the actual values were calculated, respectively. In order to test the reliability of the model, the residuals were analyzed by kernel density estimation and autocorrelation method. The results show that the residuals are white noise at 95 % confidence interval. The temperature prediction model can not only deepen people's understanding of future climate change but also has a certain reference value for both the meteorological department to carry out future temperature warnings and the environmental department to adjust the environmental protection policies.