DaveM@2: 
DaveM@2: load('Adobe.mat')
DaveM@2: load('Results1Percent.mat')
DaveM@2: 
DaveM@2: datamap = featuredata(end).IdxVar;
DaveM@2: reduceData = Data(:,datamap);
DaveM@2: reduceLabels = Labels(datamap);
DaveM@2: %% UNUSED
DaveM@2: % reduceFeatures = FeatureNames(datamap);
DaveM@2: 
DaveM@2: %% UNUSED
DaveM@2: load('Results1Percent.mat')
DaveM@2: 
DaveM@2: %% 
DaveM@2: % Use Subset of data
DaveM@2: reduceFeatures = featuredata(1).FeatureNamesRanked;
DaveM@2: 
DaveM@2: dataToUseSize = 500;
DaveM@2: dataToUse = ceil(rand(dataToUseSize,1)*size(reduceData,1))';
DaveM@2: 
DaveM@2: dMap = pdist(reduceData(dataToUse,:));
DaveM@2: clusterMethod = 'ward';
DaveM@2: % 'average'     Unweighted average distance (UPGMA)
DaveM@2: % 'centroid'	Centroid distance (UPGMC), appropriate for Euclidean distances only
DaveM@2: % 'complete'	Furthest distance
DaveM@2: % 'median'      Weighted center of mass distance (WPGMC), appropriate for Euclidean distances only
DaveM@2: % 'single'      Shortest distance
DaveM@2: % 'ward'        Inner squared distance (minimum variance algorithm), appropriate for Euclidean distances only
DaveM@2: % 'weighted'	Weighted average distance (WPGMA)
DaveM@2: 
DaveM@2: dl = linkage(dMap, clusterMethod);
DaveM@2: dendrogram(dl)
DaveM@2: incon_sp = inconsistent(dl)
DaveM@2: % figure; imagesc(squareform(dMap_sp))
DaveM@2: % title('euclidian self similarity');
DaveM@2: 
DaveM@2: %%
DaveM@2: % Use all data
DaveM@2: 
DaveM@2: dMapAll = pdist(reduceData);
DaveM@2: clusterMethod = 'ward';
DaveM@2: % 'average'     Unweighted average distance (UPGMA)
DaveM@2: % 'centroid'	Centroid distance (UPGMC), 
DaveM@2: %                     appropriate for Euclidean distances only
DaveM@2: % 'complete'	Furthest distance
DaveM@2: % 'median'      Weighted center of mass distance (WPGMC),
DaveM@2: %                     appropriate for Euclidean distances only
DaveM@2: % 'single'      Shortest distance
DaveM@2: % 'ward'        Inner squared distance (minimum variance algorithm), 
DaveM@2: %                   appropriate for Euclidean distances only
DaveM@2: % 'weighted'	Weighted average distance (WPGMA)
DaveM@2: 
DaveM@2: dl_all = linkage(dMapAll, clusterMethod);
DaveM@2: % [~,T] = dendrogram(dl_all,0)
DaveM@2: 
DaveM@2: %%
DaveM@2: % print filelist for each cluster
DaveM@2: 
DaveM@2: numClusters = 100;
DaveM@2: fnames = cell(1,numClusters);
DaveM@2: [~,T] = dendrogram(dl_all,numClusters);
DaveM@2: plotName = ['data/ClusterWith' num2str(numClusters) 'Elements'];
DaveM@2: saveas(gcf, plotName, 'fig');
DaveM@2: saveas(gcf, plotName, 'pdf');
DaveM@2: for i = 1:numClusters
DaveM@2:     numFiles = sum(T==i);
DaveM@2:     fnames{i} = Filenames(find(T==i));
DaveM@2: end
DaveM@2: 
DaveM@2: %
DaveM@2: % makeCSV for Weka
DaveM@2: % format 
DaveM@2: 
DaveM@2: feats = reduceData;
DaveM@2: 
DaveM@2: csvOut = num2cell(feats);
DaveM@2: csvOut = [csvOut, num2cell(T)];
DaveM@2: csvOut = [[reduceFeatures(datamap)', {'Class'}]; csvOut];
DaveM@2: cell2csv(['data/wekaReducedFeaturesWithNew' num2str(numClusters) '.csv'],csvOut)
DaveM@2: 
DaveM@2: 
DaveM@2: %%
DaveM@2: % fnames to CSV
DaveM@2: 
DaveM@2: maxLen = size(fnames,2);
DaveM@2: 
DaveM@2: for i = 1:maxLen
DaveM@2:     depth = size(fnames{i},1);
DaveM@2:     for ii = 1:depth
DaveM@2:         csvOut(i,ii) = fnames{i}(ii);
DaveM@2:     end
DaveM@2: end
DaveM@2: 
DaveM@2: printString = '';
DaveM@2: for i = 1:maxLen
DaveM@2:     printString = [printString ' %s, '];
DaveM@2: end
DaveM@2: 
DaveM@2: fid = fopen('junk.csv','w');
DaveM@2: fprintf(fid,[printString '\n'],csvOut{1:end,:});
DaveM@2: % fprintf(fid,'%f, %f, %f\n',c{2:end,:})
DaveM@2: fclose(fid) ;
DaveM@2: % dlmwrite('test.csv', csvOut, '-append') ;
DaveM@2: 
DaveM@2: %%
DaveM@2: T = cluster(dl_sp,'cutoff',1.3);
DaveM@2: figure; plot(T);
DaveM@2: 
DaveM@2: 
DaveM@2: 
DaveM@2: %%
DaveM@2: 
DaveM@2: 
DaveM@2: T = cluster(dl_sp,'maxclust',2);
DaveM@2: plot(T)
DaveM@2: %%
DaveM@2: T = cluster(dl_sp,'maxclust',3);
DaveM@2: plot(T)
DaveM@2: %%
DaveM@2: T = cluster(dl_sp,'maxclust',4);
DaveM@2: plot(T)
DaveM@2: T = cluster(dl_sp,'maxclust',5);
DaveM@2: plot(T)
DaveM@2: T = cluster(dl_sp,'maxclust',6);
DaveM@2: plot(T)
DaveM@2: T = cluster(dl_sp,'maxclust',7);
DaveM@2: plot(T)
DaveM@2: T = cluster(dl_sp,'maxclust',8);
DaveM@2: plot(T)
DaveM@2: T = cluster(dl_sp,'maxclust',9);
DaveM@2: plot(T)
DaveM@2: %%
DaveM@2: T = cluster(dl_sp,'maxclust',10);
DaveM@2: plot(T)
DaveM@2: %%
DaveM@2: T = cluster(dl_sp,'maxclust',100);
DaveM@2: plot(T)
DaveM@2: %%
DaveM@2: median(T)
DaveM@2: 
DaveM@2: 
DaveM@2: T = cluster(dl_sp,'maxclust',1000);
DaveM@2: median(T)
DaveM@2: 
DaveM@2: 
DaveM@2: plot(T)
DaveM@2: csvwrite('dataOutput',reduceData);
DaveM@2: 
DaveM@2: 
DaveM@2: 
DaveM@2: 
DaveM@2: 
DaveM@2: 
DaveM@2: 
DaveM@2: 
DaveM@2: 
DaveM@2: 
DaveM@2: 
DaveM@2: 
DaveM@2: % dMap_euc = pdist(reduceData);
DaveM@2: % dMap_cos = pdist(reduceData,'cos');
DaveM@2: % dMap_cos = pdist(reduceData,'cosine');
DaveM@2: % dl_euc = linkage(dMap_euc);
DaveM@2: % dl_cos = linkage(dMap_cos);
DaveM@2: % % dl_sp
DaveM@2: % dl_sp(10,:)
DaveM@2: % dl_sp(1:10,:)
DaveM@2: % sprintf('%f', dl_sp(1:10,:))
DaveM@2: % dl_sp(1:10,:)
DaveM@2: % format short g
DaveM@2: % dl_sp(1:10,:)
DaveM@2: % plot(dl_sp(:))
DaveM@2: % plot(dl_sp(:,3))
DaveM@2: % incon_sp = inconsistent(dl_sp)