Floods caused by rainstorms occur almost every year in Tumen River Basin. The floods caused severe damage to local residents, property and the ecological environment. For reducing the damage of floods, it is necessary to produce the flood susceptibility map. For mapping the flood prone areas, four algorithms such as genetic algorithm (GA), particle swarm optimization (PSO), artificial fish swarm algorithm (AFSA) and artificial bee colony algorithm (ABC) were used to optimize back propagation (BP) neural network respectively. The accuracy of the model was evaluated by Receiver Operating characteristic (ROC), root mean square error(RMSE) and mean absolute error (MAE). A total number of 222 flooded and un-flooded locations were identified. 155 locations of which were randomly selected to training the model, and the remaining 67 locations were used to validate the model. A total of thirteen flood conditioning factors，including altitude, slope, aspect, curvature, Topographic Wetness Index (TWI), Stream Power Index (SPI), Sediment Transport Index (STI), distance to river, landuse, rainfall, lithology, runoff coefficient and soil type were used in the proposed models. The area under the curve (AUC) obtained from ROC indicated that AFSA-BP showed a high accuracy (99.91%), and ABC-BP had the lowest prediction accuracy (97.28%). According to the generated flood susceptibility map, about 13% of Tumen River Basin is under very high flood susceptibility. The distance to river, rainfall, altitude and land use have the greatest impact on flood susceptibility. The results showed that the BP model optimized by the four algorithms is an effective and reliable tool for mapping the flood susceptibility. The flood susceptibility map can be used in flood prevention strategies.