A landslide stability calculation method based on Bayesian network

The calculation method of landslide stability is a critical issue to landslide research. Because the landslide is an unbalanced and unstable complex system, meanwhile the interactions among the various factors that composed a landslide are uncertain and random, the landslide stability calculation method based on probability becomes the future trend. This paper presents a new landslide stability calculation method based on Bayesian Network, which applies K2 algorithm and Bayesian method to learn Bayesian network´s structure and parameters. In the established Bayesian network model a joint tree inference algorithm is used to analyze and calculate the landslide stability under the effects of slope height, slope angle, bulk density, angle of internal friction, cohesion and etc. The paper uses 5-fold Cross-Validation to verify the accuracy in the proposed model. Compared with the calculation method based on Support Vector Machine (SVM), its reliability is higher and prediction results are better. The proposed model can directly represent the interaction mechanism between the decision-making behavior and effect factors.