Flood vulnerability and buildings’ flood exposure assessment in a densely urbanised city: comparative analysis of three scenarios using a neural network approach