Abstract: Affected by the operation of the Three Gorges Reservoir, sediment amount entering the Jingjiang reach of middle Yangtze River was sharply decreased, resulting in intensive river bed incision and bank erosion, which would not only affect the channel stability, but also increase the flood control pressure.In this paper, firstly 10 bank collapse influencing factors were selected from the aspects of water-sediment conditions and riverbed boundary conditions, and quantified.Secondly, based on the measured data, the random forest model was used to calculate the possibility of bank collapse, and the model was calibrated and verified.Finally, based on the DS evidence theory, three early-warning indicators(bank erosion possibility, nearby residential area and the width of floodplain) were combined to divide the early-warning level of bank collapse of typical sections.The results showed that:(1)the random forest model could well reflect the correlation between the influencing factors of bank collapse and its occurrence possibility, with a high accuracy of 75% being obtained during the model test.(2)The obtained early-warning levels agreed with the published results of Changjiang Water Resource Commission, and especailly at the right banks of sections of Jing 60L and Jing 98R,the calculated warning levels of bank collapse was relatively high, with grade I or grade II from 2018 to 2020.Furthermore, analysis on the contributions of different influencing factors to bank collapse in the Jingjiang reach indicated that bank revetment had the greatest contribution of 16.6%,followed by the bank slope of 14.4%,whereas the contribution of discharge is 10.2%.