A Study of Climate Factors Based on DBSCAN Clustering Algorithm and ARIMA Time Series

This study aims to systematically analyze the impact of climate factors on global temperature change by combining the DBSCAN clustering algorithm with the ARIMA time series model. First, the temperature data of each city were clustered using the DBSCAN algorithm, and the cities were divided into four categories for in-depth analysis. The temperature trends of different cities were demonstrated through visualization techniques to confirm the upward trend of global temperatures. In addition, using the MK mutation test, it was found that the temperature rise did not pass the significance test, indicating that the phenomenon was not beyond the historical fluctuation range. In order to predict future temperature changes, Elman neural network, BP neural network and ARIMA time series models were developed and Elman neural network was found to have high prediction accuracy by comparison. Further least-squares regression analysis shows that global temperature has an upward trend over time. In addition, this study examines the effect of latitude on temperature and finds that latitude has a significantly greater effect on temperature than longitude. Multiple climatic factors are synthesized and identified as the major contributors to the increase in temperature. Finally, it is expected to promote public awareness of climate warming and the implementation of countermeasures.