The LS-SVM modeling method is founded on a
relatively solid statistical learning theory. Generally,
LS-SVM can obtain a good prediction performance with the
introduction of SRM principle, kernel function and
regularization technique. Besides, LS-SVM transforms the
inequality type constraints into equalities, avoiding the
solution of the quadratic programming problem which is very
time consuming [11]. This improvement greatly reduces the
computational load and makes the LS-SVM method have
increasing applications.