With the increase of economic development and the frequency and intensity of flood disasters, post-disaster emergency management requires rapid understanding of disaster losses. This paper firstly constructed an index system from five aspects including disaster-causing factors, disaster-bearing bodies, disaster-pregnant environment, emergency response capacity and disaster situation, and verified its rationality based on generalized gray correlation analysis. Secondly, Gaussian process regression model was introduced to pre-evaluate and simulate the economic losses of flood disasters. Finally, the method was applied to evaluate the direct economic losses of flood disasters in Beijing-Tianjin-Hebei urban agglomeration from 2010 to 2020. The results show that the generalized gray correlation analysis-Gaussian process regression model has the best fitting accuracy when comparing the simple Gaussian process regression with the neural network assessment model.