Based on the consideration of the risk of disaster factors, the sensitivity of the disaster environment, and the vulnerability of the disaster bearing body, a flood risk assessment index system for the Xiaoqing River Basin was constructed. An improved AHP-entropy weight method based on D-S evidence theory was proposed to calculate the index weight so as to obtain the flood risk index. Then, the natural breakpoint grading method was adopted to determine the flood risk level and the spatial distribution of flood risk in the Xiaoqing River Basin. The results show that the overall risk of flood disasters in the Xiaoqing River Basin showed a trend of low in the south but high in the north, with high risk areas and higher risk areas accounting for 8.7% and 14.3% of the basin area, respectively, which mainly distributed on toth banks of the main stream and the main tributaries of the Xiaoqing River. The evaluation results obtained are consistent with the actual distribution of flood risk during the occurrence of Typhoon Lekima, proving that the improved method is superior to individual AHP and entropy weight method. The research results can provide a technical support for the decision-making of flood control and disaster reduction in the Xiaoqing River Basin.