// //SVM support vector machine model command. // //syntax with option svm [segment variable] by [explaining variables]/ options ; /** Options type select kernel function liner :liner function u'*v + coef mult :polynomial function (u'*v + coef)**degree gauss :gaussian functio exp(-delta*|u-v|^2)(default) laplase:laplase function exp(-delta*|u-v|) sigmoid:sigmoid function tanh 1/exp(1+exp(-*u'*v+coef)) margin=.xx //soft margin 0 < .xx is hard mergin svm (default = 1000.0) nu=.xx //nu conditon parametar (default=0.5) delta=.xx //see kernel function at gamma (default=0.1 k:dimension of data) degree=nn //see kernel funciton at nn (default=4) test=.yy //.yy is rate of data record for inspection of model. 0.0 < .yy <= 1.0 (defalut=0.0) nu = .xx //.xx is nu used by nusvm (default=0.5) coef = zz // zz is constant coeficent for kernel (default=0.0) **/ //Example 1 hand kbn x y/ 1 0.019369357 0.526381373 1 0.386283708 0.347418479 -1 0.903627442 0.208211454 -1 0.736431183 0.056018726 -1 0.701974639 0.371235275 1 0.304254514 0.112965579 1 0.29491199 0.646280255 -1 0.541982287 0.3297108 1 0.382667239 0.301098282 -1 0.802787022 0.29755779 ; attr name type/ kbn code ;
//View point and supervised label (1 or -1) plot scat x y by kbn;
//Execute svm model by kernel function liner. svm kbn by x y/ type=liner ;
No.2 is incorrect recognition,so correct column is 2.
//get output of svm for inspection get freq@ana; //Draw graph point with svm value which shows separated space. plot scat x y svm;
//Draw graph points by correct flag. plot scat x y by correct;
supervised 1 and predict 1 then correct column is 1
supervised -1 and predict -1 then correct column is -1
supervised 1 and predict -1 then correct column is 2
supervised -1 and predict 1 then correct column is 3