% below are the load commands for different databases
% depending on which database you would like to be processed
% just uncomment the related lines


allMyDatasets = {'coc81_1_1','coc81_1_1_organic','coc81_1_1_embedded','ISBSG_Banking_Applications','SDR_Banking','COCOMO_NASA93','nasa93_center_5','nasa93_center_2.csv','albrecht','COCOMOII_SDR','desharnais'};

for datasetCounter = 1:10
    eval(['load ',char(allMyDatasets(datasetCounter)),'.csv;']);
    eval(['dataset = ',char(allMyDatasets(datasetCounter)),';']);
    eval(['clear ',char(allMyDatasets(datasetCounter)),';']);
    % load coc81_1_1.csv;
    % dataset = coc81_1_1;
    % clear coc81_1_1;
    
    
    % define the variables to keep tract of.. well pretty much everything
    selectedBestLevels = -1 * ones(20,size(dataset,1));
    varianceOfSelectedBestPlaces = -1 * ones(20,size(dataset,1));
    varianceOfSelectedBestPlacesGAC2 = -1 * ones(20,size(dataset,1));
    %selectedBestKValues = -1 * ones(20,size(dataset,1));
    
    actuals = -1 * ones(20,size(dataset,1));
    actualsWithFeatures = [];
    
    pred0 = -1 * ones(20,size(dataset,1));
    mre0 = -1 * ones(20,size(dataset,1));
    
    predGac2 = -1 * ones(20,size(dataset,1));
    mreGac2 = -1 * ones(20,size(dataset,1));
    instanceSizeGac2 = -1 * ones(20,size(dataset,1));
    
    pred1 = -1 * ones(20,size(dataset,1));
    mre1 = -1 * ones(20,size(dataset,1));
    pred3 = -1 * ones(20,size(dataset,1));
    mre3 = -1 * ones(20,size(dataset,1));
    pred5 = -1 * ones(20,size(dataset,1));
    mre5 = -1 * ones(20,size(dataset,1));
    pred7  = -1 * ones(20,size(dataset,1));
    mre7 = -1 * ones(20,size(dataset,1));
    pred9 = -1 * ones(20,size(dataset,1));
    mre9 = -1 * ones(20,size(dataset,1));
    predx = -1 * ones(20,size(dataset,1));
    mrex = -1 * ones(20,size(dataset,1));
    predFloatingK = -1 * ones(20,size(dataset,1));
    mreFloatingK = -1 * ones(20,size(dataset,1));
    floatingKValues = -1 * ones(20,size(dataset,1));
    
    % a dummy to keep track of whether the best k value was found
    bestKFound = 0;
    % find the best k value for the train data
    if bestKFound == 0
        myBestK = bestk(dataset);
    end
    
    % repeat 20 times
    for counter = 1:20
        
        % randomize dataset
        dataset = randomizeDataset(dataset);
        
        % for each row in the randomized dataset do the following
        i = 0;
        while i < size(dataset,1)
            % increment i by 1
            i = i + 1;
            
            % pick up the row
            myRow = dataset(i,:);
            
            % record it into actuals
            actuals(counter,i) = myRow(1,size(myRow,2));
            % record also the features of that instance (since a lot of instances
            % have the same effort value, I needed to record the features too)
            actualsWithFeatures = [actualsWithFeatures;myRow(1,:)];
            
            
            % train is the dataset minus the selected row
            train = dataset;
            train(i,:) = [];
            
            % now start predictions
            % the one below -treeK- is our guy to defend
            [pred0(counter,i), mre0(counter,i),selectedBestLevels(counter,i),varianceOfSelectedBestPlaces(counter,i), gac2Tree, gac2Root] = treeK(myRow,train);
            
            % at this point we have build our GAC tree -above executable line- and we can check the
            % instances with the second tree
            
            [predGac2(counter,i), mreGac2(counter,i),instanceSizeGac2(counter,i),varianceOfSelectedBestPlacesGAC2(counter,i)] = secondGac(gac2Tree, gac2Root, myRow);
            if predGac2(counter,i) == -1    % meaning there were not enough instances for gac2 tree
                i = i-1;
                continue;
            end
            
            % below are the ones for various k values
            [pred1(counter,i), mre1(counter,i)] = nnk(myRow,train,1);
            [pred3(counter,i), mre3(counter,i)] = nnk(myRow,train,3);
            [pred5(counter,i), mre5(counter,i)] = nnk(myRow,train,5);
            [pred7(counter,i), mre7(counter,i)] = nnk(myRow,train,7);
            [pred9(counter,i), mre9(counter,i)] = nnk(myRow,train,9);
            %         floatingKValues(counter,i) = bestFloatingK(myRow,train);
            %[predFloatingK(counter,i), mreFloatingK(counter,i)] = nnk(myRow,train,floatingKValues(counter,i));
            [predx(counter,i), mrex(counter,i)] = nnk(myRow,train,myBestK);
        end
        counter
    end
    
    % save workspace
    eval(['save(teak_odd_numbered_k_experiment_workspaces\',char(allMyDatasets(datasetCounter)),');']);
    
end