You could use the stability index or factor of safety calculated by SINMAP or similar model as one of the environmental variables for a statistical or machine learning approach. One example of this approach is by Davis and Sims (2013), J. Soils Sediments 13:1784-1801. However, they found that the results differed little from simply using slope angle instead of the stability index. The problem may be that physical based models only work well at small spatial resolutions, whilst statistical models are usually applied at a coarser spatial resolution.   

thanks for your suggestion.