Abstract: The rainfall-induced landslides data in the Yucheng district，Ya’an city in Sichuang province are chosen as the sample data. Rainfall factors and geologic factor are extracted as the condition attribute of the rough sets. Based on the data mining and intelligent knowledge discovery function of the rough sets theory, the valuable rules are generated which can make judgments to rainfall-induced landslides occurrence. The decision rule can provide the prediction of regional landslides. The nonlinear characteristic and geological factors of the landslides are considered, and the method has a better predictive effect and higher spatial resolution than the traditional statistical method.