Urban flood risk mapping is an important tool for the mitigation of flooding in view of human activities and climate change. Many developing countries, however, lack sufficiently detailed data to produce reliable risk maps with existing methods. Thus, improved methods are needed that can improve urban flood risk management in regions with scarce hydrological data. Given this, we estimated the flood risk map for Rasht City (Iran), applying a composition of decision-making and machine learning methods. Flood hazard maps were produced applying six state-of-the-art machine learning methods such as classification and regression trees (CART), random forest (RF), boosted regression trees (BRT), multivariate adaptive regression splines (MARS), multivariate discriminant analysis (MDA), and support vector machine (SVM). Flood conditioning parameters applied in modeling were elevation, slope angle, aspect, rainfall, distance to river (DTR), distance to streets (DTS), soil hydrological group (SHG), curve number (CN), distance to urban drainage (DTUD), urban drainage density (UDD), and land use. In total, 93 flood location points were collected from the regional water company of Gilan province combined with field surveys. We used the Analytic Hierarchy Process (AHP) decision-making tool for creating an urban flood vulnerability map, which is according to population density (PD), dwelling quality (DQ), household income (HI), distance to cultural heritage (DTCH), distance to medical centers and hospitals (DTMCH), and land use. Then, the urban flood risk map was derived according to flood vulnerability and flood hazard maps. Evaluation of models was performed using receiver-operator characteristic curve (ROC), accuracy, probability of detection (POD), false alarm ratio (FAR), and precision. The findings showed that the CART method is most accurate method (AUC = 0.947, accuracy = 0.892, POD = 0.867, FAR = 0.071, and precision = 0.929). The results also demonstrated that DTR, UDD, and DTUD played important roles in flood hazard modeling; whereas, the population density was the most significant parameter in vulnerability mapping. These findings indicated that machine learning methods can improve urban flood risk management significantly in regions with limited hydrological data.