Data quality evaluation of multisource rainfall stations in Yujiang River Basin based on clustering algorithm

To solve the problem of redundant and insufficient construction of the rainfall station network in the Yujiang River Basin, it is necessary to evaluate the data quality of multi-source rainfall stations in Yujiang. This article uses the combination clustering algorithm of K-means and mixture Gaussian model to conduct experiments on the flood season data of 65 rainfall stations from 2020 to 2022, and compares the independent experimental effects of the two algorithms. The combination clustering algorithm draws on the advantages of two commonly used clustering algorithms of different types, and the F1 value reaches 70.97%. Through experiments, it can be seen that combination clustering methods have inherent advantages in discovering abnormal rainfall stations, making it easier to assess the data quality of rainfall stations over a longer time scale. They are efficient, highly automated, and provide a good idea for evaluating the data quality of rainfall stations.