Abstract: The prediction of landslide intensity is the foundation and the key of landslide risk analysis and control. Field survey was carried out in Heifangtai. Then mathematical statistic analysis of geometrical parameters for 35 landslides was conducted to systematically analyze the correlation between sliding distance and landslide geometric parameters. Also, both single and multiple regression statistical models for landslide spatial predictions were established on the basis of the loess landslide and the loess-bedrock landslide. In the statistical model, by taking the original slope gradient, the slump angle and the width of landslide as independent variables and the reach angle as the induced variable, landslide intensity prediction is easily achieved with the help of both single factor and multiple factors fitting methods.