<p>As the frequency of flooding increases due to the impacts of climate change driven by global warming, identifying flood-vulnerable areas has become critical for disaster response and emergency planning. This study aims to produce flood vulnerability maps for Seoul and conduct a comparative analysis of vulnerability between 2010 and 2021. To this end, datasets including inundation records, topography, geology, soil, land use, and hydrological systems were utilized. Deep Neural Network (DNN) and Convolutional Neural Network (CNN) deep learning models were employed to predict flood vulnerability. The models were trained and tested using data from 2010 to 2011. The trained DNN and CNN models demonstrated significant predictive performance, achieving Average Precision (AP) scores of 0.865 and 0.925, and Root Mean Squared Error (RMSE) values of approximately 0.270 and 0.245, respectively. Notably, the CNN model exhibited slightly superior performance compared to the DNN model. Furthermore, the trained models were applied to 2021 input data to generate a vulnerability map, which was then compared to the 2010 vulnerability map. The most significant change observed between the 2010 and 2021 input data was in the hydrological system dataset. An analysis of the change in flood vulnerability attributed to the improved hydrological system confirmed that Seoul’s overall vulnerability decreased in 2021 compared to the baseline period. This indicates that the flood vulnerability maps generated by the DNN and CNN models can provide meaningful information for policy-making and the identification of priority areas for flood mitigation management.</p>