Based on the contrast between the measured monthly mean temperature anomaly and the simulated monthly mean temperature anomaly reconstructed with multiple regression analyses at Moyu Station, Muqi Station and Akdala Station, the accuracies of the multiple regression analyses are analyze through the means of the cross-validation based on MAE and RMSIE, M-K mutation tests and wavelet analyses. Similarly, based on the analyses of EOF the contrast between the measured monthly mean temperature data from 90 weather stations and the data interpolated and reconstructed from 105 weather stations in Xinjiang also show that there is high similarity in eigenvectors. Finally, it is found that the multiple regression analysis possesses higher accuracy. The reconstructed Xinjiang monthly temperature data have reliability same as the measured monthly temperature data, which could better reflect the characteristics and variation of the monthly mean temperature and provide basic data for climate change research.