[Objective] The relation between landslide hazards and antecedent rainfall in 2008 to 2014 was analyzed. A meteorological early-warning model for the rainfall-induced landslide hazards was established and the weight coefficients of the model were optimized in order to provide an efficient reference for disaster warning.[Methods] Through statistical analysis of the antecedent rainfall based on the precipitation information, we adopted logistic regression analysis to calculate the weights of daily rainfall and optimize the weight coefficients of the model.[Results] The weights of disaster-occurring day rainfall, the 2-day and the 3-day rainfall before landslide disaster are 0.587, 0.220 and 0.189. The weight coefficients of the model for geological background and rainfall are 0.394 and 0.606. The previously occurred disasters were used to test and verify the model and the accuracy is 78.36%. The massively occurred disasters were also taken in July 10, 2013 as a single example to verify the model and the result is rather good.[Conclusion] The model has good early warning effect, so as to provide an efficient reference for the disaster prevention and mitigation.