Predictions and Research about Global Warming Based on ARIMA models

Abstract

Initially, by processing the data of global average temperature change from 1850 to 2022 and comparing it with the temperature data of March 2022, the data were processed to get the preliminary conclusion of data visualization and an ARIMA model and an SVM model were built by doing the time series in this way.This comparative model can predict the global average temperature at different times in the future and the time to reach 20°C in different regions. Also, the errors MSE, MAE, R2, and EV of the two models were compared by short-term predictions and the ARIMA model was found to be better. To further improve the model, Pearson correlation coefficients of global temperature with time and geographical location were introduced, and it was found that there was no necessary connection between temperature and time and longitude. Considering the negligible effect of Earth's rotation and revolution on temperature, the weak relationship between latitude and temperature is also negligible. According to the literature[1], the outbreak of natural disasters is directly related to the accumulation of greenhouse gases, so greenhouse gas data are used to train the ARIMA and SVM models, which in turn leads to a Pearson correlation coefficient of 0.6 - 0.7 between greenhouse gases and temperature. Synthesizing the above data and literature, it can be concluded that the main factor of global warming is the irrational use of natural resources by human beings, which leads to the disturbance of the earth's ecological environment and the serious deterioration of the greenhouse effect.