Abstract: The geological environment in southeast Tibet is complicated and disasters occur frequently. Landslides pose a great threat to engineering construction and human and financial safety in the region. In order to select models with higher precision for regional landslide prediction in southeast Tibet, this paper used modified landslide point data through field investigation, combined with topographic and geomorphic factors, geological factors, land cover factors and induced factors, and screened the factors through principal component analysis. Frequency ratio models, BP neural network models, and a combination of the two models were used for regional landslides prediction in southeast Tibet. Finally, ROC curves were used to evaluate the model accuracy. The results showed that the frequency ratio model after factor selection had the highest prediction accuracy for southeast Tibet (AUC=0.889). Models with factors removed through principal component analysis had higher accuracy than those without removal, and landslides in southeast Tibet were mainly distributed along river systems, including the Yarlung Zangbo River, Daqu River, Zangqu River, Nujiang River, Lancang River, Weiqu River, Janqu River and Zhaqu River. The models were used to predict the disaster in the study area, revealing that landslide points were located in high susceptibility and susceptibility zones. The models developed in this study can provide technical reference for engineering construction in southeast Tibet.