Abstract: With the development of in-depth research, spatial prediction of perilous rock has become the focus and difficult point. The key function of geo-hazard spatial prediction is to express the danger degree under different geologic conditions, which is equivalent to hazard degree assessment. Current research methods in this field are mainly qualitative, semi-quantitative or depending on experts′ experience. This article has set up a quantitative index system based on information entropy method, subjective weight of all indices are combined with data from 37 perilous rocks in Wanzhou County, Three Gorges, and a qualitative perilous rock hazard degree assessment model was established. Then this model was used in evaluating the hazard degree of the perilous rocks in Wanzhou County, danger degree result of them are obtained, and verified through field investigation, it’s indicated that accuracy reaches 86.5%. Consequently this model can be implemented as a new method for risk evaluation or spatial prediction of perilous rocks, and thus provide a basis for early warning.