Based on the existing researches, and in combination with the knowledge of artificial intelligence, management decision science, and theoretical derivation, this study constructs a flood forecasting model based on the improved neural network algorithm, applies it to flood forecasting and flood damage assessment, proposes a flood decision support system consisting of external conversion layer, information processing application layer and database layer, and optimizes BP neural system in view of its defects of slow convergence speed and long iteration time by genetic algorithm. The actual forecasting results show that the improved BP neural network proposed in this study has higher forecasting accuracy, and the correlation coefficient between forecasting results and actual monitoring results reaches 96%, far exceeding the forecasting accuracy of RBF neural network algorithm (68%), which proves the feasibility and rapidity of the proposed algorithm. The improved BP neural network algorithm is used for rapid flood disaster assessment, and the needs of flood disaster assessment in different stages are specifically analyzed. In terms of accuracy, the RBF algorithm's disaster assessment grades are relatively small, and the assessment effect using the algorithm proposed in this paper is the best.