Abstract: Evaluation of landslide susceptibility provide an importance evidence of landslide disaster prevention and landslide warning. A new combined weight process(RF-AHP) based on random forest(RF), analytic hierarchy process(AHP) and distance functions is proposed. The model is applied to the northern mountainous area of Guide County. The compatibility and applicability of three models are discussed. The results show that (1) the RF-AHP model accounts for 38.38% of the high risk area and the extremely high risk area, it includes 60.13% of the landslide disasters, is greatly improved compared to the two models of AHP and RF. (2)With the gradual increase in risk levels, the actual rate of the corresponding disasters in the results of RF-AHP also gradually increases. The objective causes of the difference in results between the three methods were discussed. It can be concluded from the case study that combined weight process is suitable for evalution of landslide susceptibility.