<p>Flood disaster constitutes a major threat to urban safety, and identifying and quantifying the causal relationships between flood disaster loss and influencing factors is critical for disaster prevention and mitigation strategies. In this study, a flood disaster loss risk evaluation and attribution analysis method based on the Bayesian Network structure learning algorithm is constructed. Taking the urban agglomeration in the middle Yangtze River as the study area, the risk levels of different cities and their uncertainties are assessed, the causal relationships between disaster-inducing factors, disaster-pregnant environments, disaster- affected bodies, and disaster loss risks are derived, and the significance of the factors affecting different risk levels is analyzed. The results show that the flood disaster loss risk is higher in the Poyang Lake and Dongting Lake basins, and the risk level and its uncertainty had the same trend in general; however, there were certain exceptional area that need attention, such as Nanchang and Xiangtan with high risk and low uncertainty, Wuhan and Xiangyang with low risk and high uncertainty; moreover, the main impacting factors of different risk levels were different, e.g., the main factors for level 1 were cumulative rainfall, river density and elevation standard deviation, while the main factors for level 2 were population density, cumulative rainfall, and urbanization.</p>