evaluateoutput.m

有监督自组织映射-偏最小二乘算法(A supervised self-organising map–partial least squares algorithm),可以用语多变量数据的回归分析

function [MLKOut] = EvaluateOutput(Ytr,YpredTrain,Ytest,YpredTest,Imap,MLK,MLKP);

MLKOut=MLK;
if (upper(MLKP.ProblemType) == 'CLS' & Imap == 1)
    % classification
    if (upper(MLKP.SilentMode) == 'N')
        [ClassTrain, PercCorrTrain] = Classify(Ytr,YpredTrain,MLKP)
        [ClassTest, PercCorrTest] = Classify(Ytest,YpredTest,MLKP)
    else
        [ClassTrain, PercCorrTrain] = Classify(Ytr,YpredTrain,MLKP);
        [ClassTest, PercCorrTest] = Classify(Ytest,YpredTest,MLKP);
    end
    MLKOut.ClassTrain=ClassTrain;
    MLKOut.ClassTest=ClassTest;
    MLKOut.PercCorrTrain=PercCorrTrain;
    MLKOut.PercCorrTest=PercCorrTest;
elseif (upper(MLKP.ProblemType) == 'REG' | Imap == 2)
    % regression
    [Nobj,Nvar]=size(Ytr);
    [RmseTrain, CorrTrain] = Rmse(Ytr,YpredTrain);
    RmseTrainAll = sqrt(RmseTrain*RmseTrain'/Nvar);
    [RmseTest, CorrTest] = Rmse(Ytest,YpredTest);
    RmseTestAll = sqrt(RmseTest*RmseTest'/Nvar);
    if (upper(MLKP.SilentMode) == 'N')
        for ivar=1:Nvar
            Message=sprintf('RMSE training: %g   Corr:  %g', RmseTrain(ivar), CorrTrain(ivar));
            disp(Message);
            Message=sprintf('RMSE test:     %g   Corr:  %g', RmseTest(ivar), CorrTest(ivar));
            disp(Message); disp(' ');
        end
    end
    MLKOut.RmseTrain=RmseTrain;
    MLKOut.RmseTest=RmseTest;
    MLKOut.CorrTrain=CorrTrain;
    MLKOut.CorrTest=CorrTest;
    MLKOut.RmseTrainAll=RmseTrainAll;
    MLKOut.RmseTestAll=RmseTestAll;
end