Abstract: Prediction of typhoon-induced debris flow can help to reduce casualties and losses. This study focused on 47 debris flow catchments of the Chenyulan River Watershed in Nantou County,Taiwan,China. Seven variables-mean channel slope,effective drainage area,shape coefficient,channel length,lithology,area ratio of collapses and landslides,and mean rainfall intensity-were selected from the conditioning factors of rainfall-induced debris flow,including geomorphology,loose debris materials and rainfall conditions. According to the importance of these variables,the area ratio of collapses and landslides as well as the mean rainfall intensity were used to construct the predictive model. Random sampling selection was used to divide the dataset into the training and validation datasets (70% and 30% of the dataset,respectively). The former was used to develop the predictive model and the latter to validate the model. Four indicators,namely precision,accuracy,false negative rate,and false position rate,were employed as indexes of quantitative assessment and used to determine the optimal model. The results show that the prediction model of typhoon-induced debris flow based on Fisher discriminant analysis has good prediction performance,and it compensated for the deficiency of rainfall intensity-duration thresholds model that analyze critical rainfall of debris flow based only on rainfall data. A comparison of the predictions of the I-D model and our model indicated that ours had superior prediction performance. The results provide strong technical support for the prediction of rainfall-induced debris flow.