Abstract: This paper investigates the sensitive factors of the average speed of debris flows and coupling relationship of the influence factors with the data of debris flows in Jiangjia gully. The BPANN and SVM are proposed for building predictive model of average velocity of debris flows. Prediction accuracy and scope of application of the two models are compared. Analysis is done on various factors for their sensitivities to average speed of debris flows with the default factor method. A prediction model for debris flows is established and verified. The results show that the prediction accuracy of the SVM model is better than that of the BPANN model in the validation phase. The extrapolating ability and predicting capability have been validated by using the support vector machine prediction model,and the SVM model is more suitable for its prediction of debris flow. Channel slope and the thickness of the unstable layer are the main sensitive factors affecting average velocity of debris flows. The interaction amongst various factors formed coupling relationship under the complicated condition. The relationship model of average speed of debris flows established base upon channel slope,and the thickness of the unstable layer can well express quantitatively the response relationship between debris flows dynamic and various factors with sufficient high accuracy. It is an accurate scientific basis for prevention of debris flow.