It extracted learning samples from the MADM problem,estimated the Multiple Attribute Utility Function via LSSVM,and then selected the most desirable alternative.