In this paper, through combining information diffusion principle and BP neural network theory, a new prediction model of drought disaster assessment is established. First, the original data are fuzzily processed based on information diffusion method, then a new training sample is formed; second, the new sample is used to design and train BP neural network; finally, the trained fuzzy neural network may evaluate drought disaster. The model is applied in the Haihe River basin, the results show the model can resolve the phenomenon of fewer samples and the existing contradictory samples, good results have been achieved for drought disaster assessment.