Abstract: The K-PSO clustering algorithm and entropy method were introduced to establish a sensitivity analysis model for landslide. The 22 typical landslides located in Xulong hydropower station reservoir area were investigated. Eight major factors including rock mass structure， slope structure， fault distance， signs of deformation， slope height， average gradient， induced earthquake and submerged ratio were determined for landslide sensitivity analysis. The weights of major factors determined by the entropy method are 0.152， 0.178， 0.035， 0.106， 0.106， 0.169， 0.193， 0.061， respectively. Sensitivity analysis results based on K-PSO clustering algorithm showed that among the 22 landslides， 8 landslides are evaluated as low sensitive， 9 as moderate， 4 as severely sensitive and one as extremly sensitive. Compared with the in-situ observations， the evlauated level of sensitivity of the 22 landslides agree very well with the actual development of the landslides. The proposed K-PSO method is effective for landslide sensitivity analysis in Xulong hydropower station reservoir area.

Key words: entropy method, landslide, K-PSO, clustering model, sensitivity