Keep Headers in Loop and Skip Errors

Hi. I have two issues with my loop, and I'd appreciate any help.

First I am creating excel files for impOOB results. However, I just get the values. I would like to keep the Variable Names or Header Names.

Second, some of the spreadsheets in the loops do not have all the predictors I am specifying: 'Manhole','Catch','BackUp','PRCP', 'Street'

Matlab then stops running the loop. How can I get Matlab to skip over the errors and just ignore the spreadsheets?

Finally, this is just an extra question. I really rather have all the results in one excel file with a row for each spreadsheet name and the results in columns. Is there a way I can do this?

I've included three sample files that were in my directory. (My directory actually has 174 such files). Thanks again.

indir = 'C:\Users\Desktop\FilesforRF';            %path to input directory
outdir = 'C:\Users\Desktop\FilesforRF\results';
outdir2 = 'C:\Users\Desktop\FilesforRF\results2';

dinfo = dir(fullfile(indir, '*.xlsx'));
filenames = {dinfo.name};
nfiles = length(filenames);

for K = 1 : nfiles
    thisfile = filenames{K};
    infile = fullfile(indir, thisfile);
    outfile = fullfile(outdir, thisfile);
    outfile2 = fullfile(outdir2, thisfile);
    p = readtable(infile,'PreserveVariableNames',true)   
    dsa = p;  
    X=dsa(:,ismember(dsa.Properties.VariableNames, {'Manhole','Catch','BackUp','PRCP'}))
    Y=dsa(:,ismember(dsa.Properties.VariableNames, {'Street'}))
    t = templateTree('NumVariablesToSample','all',...
    'PredictorSelection','interaction-curvature','Surrogate','on');
rng(1); % For reproducibility
Mdl = fitrensemble(X,Y,'Method','Bag','NumLearningCycles',200, ...
    'Learners',t);
yHat = oobPredict(Mdl);
R2 = corr(Mdl.Y,yHat)^2
impOOB = oobPermutedPredictorImportance(Mdl);
outdata = impOOB;
    outdata2 = R2 
    
    writematrix(outdata, outfile);       %needs R2019b or later  %the sub-table!
    writematrix(outdata2, outfile2);
end

Answer 1

[CMatlabWold]

I figured out the first two. For the last, I believe I will use an excel merge type of function

I added Mdl.PredictorNames. And for my error issues, I used try catch.

outtable = array2table(outdata,'VariableNames',Mdl.PredictorNames)

for K = 1 : nfiles
  try
    thisfile = filenames{K};
    infile = fullfile(indir, thisfile);
    outfile = fullfile(outdir, thisfile);
    outfile2 = fullfile(outdir2, thisfile);
    p = readtable(infile,'PreserveVariableNames',true)   
    dsa = p;  
    X=dsa(:,ismember(dsa.Properties.VariableNames, {'Manhole','Catch','BackUp','PRCP'}))
    Y=dsa(:,ismember(dsa.Properties.VariableNames, {'Street'}))
    t = templateTree('NumVariablesToSample','all',...
    'PredictorSelection','interaction-curvature','Surrogate','on');
rng(1); % For reproducibility
Mdl = fitrensemble(X,Y,'Method','Bag','NumLearningCycles',200, ...
    'Learners',t);
yHat = oobPredict(Mdl);
R2 = corr(Mdl.Y,yHat)^2
impOOB = oobPermutedPredictorImportance(Mdl);
impOOB(impOOB<0) = 0;
impOOB = impOOB./sum(impOOB)
outdata = impOOB;
    outdata2 = R2;
    outtable = array2table(outdata,'VariableNames',Mdl.PredictorNames)
    outtable2 = array2table(outdata2)
    writetable(outtable, outfile);       %needs R2019b or later  %the sub-table!
    writetable(outtable2, outfile2);
  catch
     fprintf('Inconsistent data in iteration %s, skipped.\n', K);  

  end
end

Unrecognized function or variable 'nfiles'.