Abstract: Flash flood disaster is one of the natural disasters that have a significant impact on human beings, causing enormous damage to the national economy and people’s lives and properties. Conducting a risk assessment of flash floods is an effective way to defend against them, and an accurate assessment of flash flood risk can provide strong technical support for flash flood prevention and decision-making. A study is conducted on the western region of Nanyang City based on 415 historical flash flood disaster points, and 12 risk indicators such as elevation, slope, mean annual rainfall, and mean annual maximum 3-hour rainfall are selected to construct a LightGBM flash flood risk assessment model. Additionally, random forest（RF） and extreme gradient boosting（XGBoost） methods are combined, and the accuracy and precision of the models are compared by using five main evaluation indicators. The model with the best performance is used to create a flash flood risk map for the study area, and a multi-scale flash flood risk assessment map is generated based on this. The distribution characteristics of flash flood disasters in the western region of Nanyang City are explored, and the causes of flash flood are analyzed. The results show that:（1） The LightGBM model performs the best with an accuracy of 0.915 7. The RF model performs poorly with an accuracy of 0.846 2；（2） slope, elevation, and mean annual precipitation are the main risk indicators affecting flash floods in the study area；（3） The flash flood risk assessment results for the study area are consistent with the actual occurrence of flash floods. The high-risk and extremely high-risk areas account for 21%of the total area and are mainly distributed near mountains, the lowest areas around rivers, and regions with high agricultural productivity.LightGBM is an effective method for assessing flash flood risks and can provide guidance for flash flood prevention and management planning in Nanyang City.