In this paper,BP neural network and random forest regression algorithm are used to correct the temperature profile data of FY-3C GNSS radio occultation in 2017.The results show that both methods can correct FY-3C radio occultation temperature data,but the performance of the random forest regression is better than that of the neural network.For the random forest regression algorithm and the neural network,the mean absolute errors between the corrected results and the reanalysis data are 0.03 K and 0.32 K,respectively,and the mean square errors are 0.09 K² and 1.02 K²,respectively.When the globe is divided into 324 grids of 10°×10°,the random forest regression algorithm yields positive returns of 97.53% and 92.9% for average absolute error and mean square error corrections,respectively,and neural network produces positive returns of 75.61% for average absolute error correction and 67.9% for mean square error correction.