DaveM@25: function [linkList, featureList]= treeLinkFeatures(data, depthThresh, featureNames)
DaveM@28: %% [linkList, featureList]= treeLinkFeatures(data, depthThresh, featureNames)
DaveM@10: % given a dataset, a hierarchical cluster of the data is produced, and then
DaveM@10: % the data is traversed, such that, for each split in the data, a set of
DaveM@10: % features are produced, which are the ranked features that can be used to
DaveM@10: % separate the given dataset at that point.
DaveM@28: % data is the nxm matrix of content, n is the number of samples and m is
DaveM@28: % the number of features.
DaveM@28: % depthThresh is a list of the range of tree depths to traverse from the
DaveM@28: % aglomerative clustering tree. A single value of depthThresh, will assume
DaveM@28: % 1:depthThresh. For analysis of a single layer of the tree, pass a list of
DaveM@28: % two values, both of which are the layer to be analysed.
DaveM@28: % feature names is the list of features, so that grown trees have suitable
DaveM@28: % names. No feature names will result in the feature number being returned.
DaveM@28: % featureList corresponds to the rows in linkList, with the form column 1
DaveM@28: % is the 5 most relevant features, column 2 is the depth and column 3 is a
DaveM@28: % decision classification tree for the decision - perhaps this should be in
DaveM@28: % the form of a struct instead?
DaveM@28: 
DaveM@10: 
DaveM@9: 
DaveM@25: if(nargin < 3)
DaveM@25:     featureNames = 1:size(data,2);
DaveM@25: end
DaveM@16: if(nargin < 2)
DaveM@19:     depthThresh = 999;
DaveM@16: end
DaveM@27: 
DaveM@31: if (length(depthThresh) == 1)
DaveM@27:     depthThresh = 1:depthThresh;
DaveM@27: end
DaveM@27:     
DaveM@9: linkList = aglomCluster(data);
DaveM@16: linkList = depthCheck(linkList);
DaveM@10: listSize = size(data,1);
DaveM@9: 
DaveM@10: % linkList(:,4) = 0;
DaveM@24: featureList = cell(listSize-1,3);
DaveM@10: currentRow = [2*listSize-1];
DaveM@9: 
DaveM@12: %%
DaveM@15: while (~isempty(currentRow))
DaveM@16:     if(currentRow(1) > listSize)
DaveM@30:         row = currentRow(1) - listSize
DaveM@17: %         rD = linkList(row,4);
DaveM@27:         if any(linkList(row,4)==depthThresh)
DaveM@16:             classList = traceLinkageToBinary(linkList, row);
DaveM@23:             featureList{row,1} = rfFeatureSelection(data(classList>0,:), classList(classList>0));
DaveM@17:             featureList{row,2} = linkList(row,4);
DaveM@24:             featureList{row,3} = fitctree(data(classList>0,featureList{row,1}),classList(classList>0),'PredictorNames',featureNames(featureList{row,1}));
DaveM@16:         end
DaveM@17:         currentRow = [currentRow; linkList(row,1); linkList(row,2)];
DaveM@10:     end
DaveM@17:     currentRow = currentRow(2:end);
DaveM@29:     save('partialResults.mat');
DaveM@9: end
DaveM@9: 
DaveM@9: end