示例#1
0
 public String mapRegModelRegressionNormalization(String method, String arg) {
   if (method == null || REGRESSIONNORMALIZATIONMETHOD.NONE.value().equals(method)) {
     return arg;
   } else if (REGRESSIONNORMALIZATIONMETHOD.EXP.value().equals(method)) {
     return "Math.exp( " + arg + " )";
   } else if (REGRESSIONNORMALIZATIONMETHOD.SOFTMAX.value().equals(method)
       || REGRESSIONNORMALIZATIONMETHOD.LOGIT.value().equals(method)) {
     return "1.0 / ( 1.0 + Math.exp( -" + arg + " ) ) ";
   } else {
     throw new UnsupportedOperationException(
         "Regression models can't support "
             + method
             + ", check that a classification model was not required instead. ");
   }
 }
示例#2
0
 public String mapRegModelClassificationNormalization(String method, String arg) {
   if (method == null || REGRESSIONNORMALIZATIONMETHOD.NONE.value().equals(method)) {
     return arg;
   } else if (REGRESSIONNORMALIZATIONMETHOD.EXP.value().equals(method)) {
     return "Math.exp( " + arg + " )";
   } else if (REGRESSIONNORMALIZATIONMETHOD.SOFTMAX.value().equals(method)) {
     return "Math.exp( " + arg + " )";
   } else if (REGRESSIONNORMALIZATIONMETHOD.LOGIT.value().equals(method)) {
     return "1.0 / ( 1.0 + Math.exp( -" + arg + " ) )";
   } else if (REGRESSIONNORMALIZATIONMETHOD.PROBIT.value().equals(method)) {
     return "probitPhi( " + arg + " )";
   } else if (REGRESSIONNORMALIZATIONMETHOD.CLOGLOG.value().equals(method)) {
     return "1.0 - Math.exp( - Math.exp( " + arg + " ) )";
   } else if (REGRESSIONNORMALIZATIONMETHOD.LOGLOG.value().equals(method)) {
     return "Math.exp( - Math.exp( -" + arg + " ) )";
   } else if (REGRESSIONNORMALIZATIONMETHOD.CAUCHIT.value().equals(method)) {
     return "0.5 + Math.atan( " + arg + " ) / Math.PI";
   } else {
     throw new UnsupportedOperationException("Unknown normalization method :" + method);
   }
 }