Exploration of Future Temperature Analysis Based on ARIMA Time Series Model and GA-BP Neural Network Prediction Model

The persistent ultra-high temperature caused by global warming effect has sounded the alarm for the development of human society. In order to improve these problems and make some constructive suggestions, this paper focuses on the main factors affecting global temperature using time series and GA-BP neural network prediction models to investigate the trend of global average temperature. We first preprocess the data using mean interpolation, and then compare the significance of the global mean temperature change using Kruskal–Wallis test with multiple independent samples concluding that the global temperature increase triggered by March 2022 does not exceed the temperature of the past decade. Then, using a time-series prediction model and a GA-BP neural network, temperatures in 2050 are predicted to be 15.275, 14.863 and 14.862,15.660, and 15.204, 15.178, based on global average annual temperature data from previous years and taking into account various factors such as carbon dioxide, solar radiation intensity and fossil fuel combustion. The above model shows that the predicted temperatures reach 20 °C only at 2758, 2912 and 2860, which implies that the global average temperature will not reach 20 °C from 2050 to 2100. Finally, we compare the performance of the above models by determining the coefficients and find that the prediction based on the GA-BP neural network which is the most accurate.