This paper presents a prediction method on the slope failure probability at mountainous area near the source of river by using both the neural network and the similarity analysis. The causative factor, i.e., geological data and remote sensing data are categorized by using similarity analysis to loosen the nonlinearity of prediction function around the boundary between different categories. The slope failure potential is computed as the probability of failure occurrence among the data which have a certain similarity each other. Finally, the relationship between the failure probability and those data in a category is fitted by the neural network system which is associated with geological basic data, remote sensing data and precipitation data. Applicability of the proposed method is discussed by using real data obtained at Shigenobu river basin.
