Abstract: Based on the principle of Mahalanobis distance discriminant analysis,a stability predictive model of landslide is established in this paper,including nine indexes reflecting the stability of landslide:configurational character,slip surface character,the shear strength of slip soils,latest activities (intrinsic factor) and potamic effect,rainstorm intensity,posterior load,engineering activities,earthquake  effect (exterior factor).Discriminant functions are obtained through training a large set of lanslide engineering in three Gorges zone.The evaluated results show that the prediction model of distance discriminant analysis excellent performance and the resubstitution rate is zero.This is a new method to predict the stability of landslide and can be used in practical lanslide engineering.

Key words: landslide, stability, prediction, distance discriminant analysis