Introduction - If you have any usage issues, please Google them yourself
The support vector m achine ( SVM ) is a prom ising artificia l inte lligence technique, in w hich the regres
sion a lgorithm has a lready been used to so lve the nonlinear function approach successfully. Un fortunate ly, m ost us
ers se lect param eters for an SVM by ru le o f thumb, so they frequently fail to generate the optim al approach ing e ffect
for the function. Th is has restricted effective use o f SVM to a great degree. In order to get optim a l param eters auto
m atically, a new approach based on an adaptive genetic a lgorithm ( AGA ) is presented, w hich autom atica lly ad
justs the param eters for SVM. This m ethod se lects crossover probability and mutation probab ility accord ing to the
fitness va lues of the object function, therefo re reduces the convergence tim e and im proves the prec ision o fGA, in
suring the accuracy of param eter se lection. Th is m ethod w as applied to m odeling of w ater level system o f a sh ip
b