Abstract: Urban flood disaster has become more frequent and serious due to the global climate change and rapid urbanization, which seriously affects the safety of people’s lives and the sustainable development of the society. Therefore, it is particularly important to study the assessment and disaster mechanism of urban flood disasters under the changing environment. There is still no widely accepted standard in flood hazard and risk assessment due to the varying perceptions and definitions of disasters by different researchers, and the flood hazard and risk assessment related indictors or methods are very uncertain and less objective. This study proposes a framework for analyzing flood disasters in Chikan District, Zhanjiang City, Guangdong Province by using k-means clustering and the Apriori algorithm. Firstly, k-means method is used to cluster each flood influencing factor. Secondly, association rule mining is used to identify the best precipitation index for evaluating flood disasters. Finally, environmental factors such as elevation, slope, impermeability and distance from the river are selected as flood influencing factors. They are combined with the best precipitation index and historical flood disaster data to extract association rules to investigate the forming mechanisms of different levels of flood disasters.Results show that flood hazards are frequent in this region, especially in urbanized areas with high impermeability, and gentle and low-lying areas are prone to more serious flood disasters. The accumulated 24-h rainfall is the most effective precipitation index in the Chikan District. When the inundation depth is 0.20～0.55 m for moderate waterlogging, the main hazard trigger is the distance from river（e.g., less than 284.61 m）, and when the inundation depth is 0.55～1.00 m for severe waterlogging, the main trigger would be the slope（e.g., less than 1.72°） and elevation（e.g.,-7～8 m）. In addition, as the most direct driving factor for the occurrence of moderate waterlogging events, the accumulated 24-h rainfall has decreased in the mechanism of severe waterlogging disasters, which indicates that the importance of environmental factors has increased in this case. These results would be helpful for providing procedures and solutions for urban flood risk management.