Additionally, several aspects should be taken into
consideration to select the suitable input variables and
establish a good LS-SVM for flooding velocity. The choice of the best sets for the LS-SVM model rests on the fulfillment of
following criteria. (1) The optimal set should lead to the best
output prediction, with a minimum prediction error. (2) The
optimal set should contain a minimum number of selected
variables. (3) The optimal set of the selected variables should
be closely associated with the phenomenology of the flooding
capacity [7-14].