Abstract:

The construction investment and spatial resolution of existing precipitation monitoring automatic meteorological observation stations, weather radars, satellites and other technologies vary greatly and are limited, resulting in differences in regional precipitation prediction accuracy and timeliness. In this study based on big data 5G communication links and precipitation data, correlation and regression analysis methods were used to explore the correlation between mobile terminal signal attenuation characteristics and precipitation. Based on the core pseudo code algorithm, linear regression, decision tree regression, and random forest regression precipitation prediction models were constructed and their performance was evaluated to improve the accuracy of precipitation prediction. The results indicated that there was a weak correlation between the communication data of the big data 5G communication link and the precipitation data. The linear regression precipitation prediction model had NSE of −0.115444, the decision tree regression precipitation prediction model had NSE of −1.065824 and the random forest regression precipitation prediction model had NSE of 0.310811. In addition, the joint data of communication and precipitation monitoring data from May to June 2022 of the big data 5G communication link in Fuzhou urban area achieved the best average prediction accuracy of 95.86% in the random forest regression precipitation prediction model. This suggests that the communication data from the big data 5G communication links can be used for high−resolution and high−precision precipitation prediction in the random forest regression precipitation prediction models. The results of this study provide a scientific alternative to high spatiotemporal resolution meteorological forecasts.

Key words: 5G communication link, Precipitation, Correlation, Regression model, Random forest