% before everything, do a clean-up of the workspace
clear;clc;

% define dataset names for different within-company like features
cocomoProjType = {'cocomo81e','cocomo81o','cocomo81s'};
nasaCenterType = {'nasa93_center_1','nasa93_center_2','nasa93_center_5'};
desharnaisLangType = {'desharnaisL1','desharnaisL2','desharnaisL3'};
chinaResource = {'chinaResource1','chinaResource2','chinaResource3','chinaResource4'};
finnishAppArea = {'finnishAppArea2','finnishAppArea7','finnishAppArea10'};
finnishAppType = {'finnishAppType1','finnishAppType2345'};
kemererHardware = {'kemererHardware1','kemererHardware23456'};
maxwellAppType = {'maxwellAppType1','maxwellAppType2','maxwellAppType3'};
maxwellHardware = {'maxwellHardware2','maxwellHardware3','maxwellHardware5'};
maxwellSource = {'maxwellSource1','maxwellSource2'};
% combine above datasets into a single variable
allMyDatasets = {...
    %     'chinaResource'...
    %     ,'finnishAppArea',...
    'cocomoProjType',...
    'nasaCenterType',...
    'desharnaisLangType',...
    'finnishAppType',...
    'kemererHardware',...
    'maxwellAppType',...
    'maxwellHardware',...
    'maxwellSource'...
    };

% open and close file for current dataset
% below is the file for number of instances
resultsFile = fopen('resultsFile.txt','w' );
fclose(resultsFile);
% below is for all win,tie,loss and actual mar, mmre, mdmre, pred25
% combined
winTieLossFileAll = fopen('winTieLossFileAll.txt','w' );
fclose(winTieLossFileAll);
% below is the file for MMRE win, tie, loss values
winTieLossFileMMRE = fopen('winTieLossFileMMRE.txt','w' );
fclose(winTieLossFileMMRE);
% below is the file for MdMRE win, tie, loss values
winTieLossFileMdMRE = fopen('winTieLossFileMdMRE.txt','w' );
fclose(winTieLossFileMdMRE);
% below is the file for Pred(25) win, tie, loss values
winTieLossFilePred25 = fopen('winTieLossFilePred25.txt','w' );
fclose(winTieLossFilePred25);
% below is the file for MAR win, tie, loss values
winTieLossFileMAR = fopen('winTieLossFileMAR.txt','w' );
fclose(winTieLossFileMAR);


% start doing cross-company and within company experiments for
% all the datasets and their subsets
for datasetCounter = 1:size(allMyDatasets,2)
    
    % load subsets of the dataset
    currentDatasetName = (allMyDatasets(datasetCounter));
    % get subset size
    eval(['subsetSize = size(' char(currentDatasetName) ',2);']);
    
    % do for each subset
    for subsetCounter = 1:subsetSize
        % define the within company and cross company datasets
        subsetIndices = [1:subsetSize];             % get indices
        wcIndex = subsetIndices(subsetCounter);     % define one as wc
        % subsetIndices(subsetCounter) = [];          % delete wc from subsets
        ccIndex = subsetIndices;                    % assing cc indices
        eval(['wcFileName = char(' char(currentDatasetName) '(' num2str(subsetCounter) '));']);
        wcDataset = csvread([wcFileName '.csv']);   % read the wc dataset
        wcDataset = myNormalizer(wcDataset);
        ccDataset = [];                             % define a var. for cc dataset
        ccDatasetSizes = zeros(1,size(ccIndex,2));
        for tmp1 = 1:size(ccIndex,2)                % append cc datasets in that for loop
            eval(['tmpName = [' char(currentDatasetName) '(' num2str(ccIndex(tmp1)) ')];']);
            tmpDataset = csvread([char(tmpName) '.csv']);
            ccDatasetSizes(tmp1) = size(tmpDataset,1);
            tmpDataset = myNormalizer(tmpDataset);
            ccDataset = [ccDataset;tmpDataset];
            % also define a variable to keep track of how many instances
            % are selected from each CC center
            eval([char(tmpName) 'Instances = -1 * ones(20,size(wcDataset,1));']);
        end
        
        % define the variables to keep track of.. well pretty much
        % everything
        % variables related to how many instances are selected from
        % different centers
        predictionZoneSizeCC = -1 * ones(20,size(wcDataset,1));
        
        % below variables keep track of stuff related to wc experiments
        pred0WC = -1 * ones(20,size(wcDataset,1));
        predGac2WC = -1 * ones(20,size(wcDataset,1));
        predictionZoneWC = -1 * ones(20,size(wcDataset,1));
        %         pred1WC = -1 * ones(20,size(wcDataset,1));
        %         pred2WC = -1 * ones(20,size(wcDataset,1));
        %         pred4WC = -1 * ones(20,size(wcDataset,1));
        %         pred8WC  = -1 * ones(20,size(wcDataset,1));
        %         pred16WC = -1 * ones(20,size(wcDataset,1));
        %         predxWC = -1 * ones(20,size(wcDataset,1));
        % below variables keep track of stuff related to cc experiments
        pred0CC = -1 * ones(20,size(wcDataset,1));
        predGac2CC = -1 * ones(20,size(wcDataset,1));
        predictionZoneCC = -1 * ones(20,size(wcDataset,1));
        %         pred1CC = -1 * ones(20,size(wcDataset,1));
        %         pred2CC = -1 * ones(20,size(wcDataset,1));
        %         pred4CC = -1 * ones(20,size(wcDataset,1));
        %         pred8CC  = -1 * ones(20,size(wcDataset,1));
        %         pred16CC = -1 * ones(20,size(wcDataset,1));
        %         predxCC = -1 * ones(20,size(wcDataset,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(wcDataset);
        %         end
        
        % randomize dataset
        wcDataset = randomizeDataset(wcDataset);
        ccDataset = randomizeDataset(ccDataset);
        
        for runCounter = 1:20  % do WC and CC experiments 20 times
            
            % for each row in the within company dataset do the following
            actualEffortValues = wcDataset(:,size(wcDataset,2));
            i = 0;
            while i < size(wcDataset,1)
                % increment i by 1
                i = i + 1;
                
                % pick up the row
                myRow = wcDataset(i,:);
                % wcTrain is the within-company dataset minus the selected row
                wcTrain = wcDataset;
                wcTrain(i,:) = [];
                
                %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
                %%%%%%%%%%%%%%%%%%%% below are predictions for within-company%%
                %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
                % now start predictions
                % the one below -treeK- is our guy to defend
                [pred0WC(runCounter,i), gac2TreeWC, gac2RootWC] = treeK(myRow,wcTrain);
                
                % at this point we have build our GAC tree -above executable line- and we can check the
                % instances with the second tree
                [predGac2WC(runCounter,i), predictionZoneWC] = secondGac(gac2TreeWC, gac2RootWC, myRow);
                if predGac2WC(runCounter,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(runCounter,i)] = nnk(myRow,wcTrain,1);
                %                 [pred2(runCounter,i)] = nnk(myRow,wcTrain,2);
                %                 [pred4(runCounter,i)] = nnk(myRow,wcTrain,4);
                %                 [pred8(runCounter,i)] = nnk(myRow,wcTrain,8);
                %                 [pred16(runCounter,i)] = nnk(myRow,wcTrain,16);
                %                 [predx(runCounter,i)] = nnk(myRow,wcTrain,myBestK);
                
            end
            
            i = 0;
            while i < size(wcDataset,1)
                % increment i by 1
                i = i + 1;
                
                % pick up the row
                myRow = wcDataset(i,:);
                % wcTrain is the within-company dataset minus the selected row
                wcTrain = wcDataset;
                wcTrain(i,:) = [];
                % ccTrain is the cross-company train dataset and it is used in
                % an as-is manner
                ccTrain = ccDataset;
                
                %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
                %%%%%%%%%%%%%%%%%%%% below are predictions for cross-company%%%
                %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
                % now start predictions
                % the one below -treeK- is our guy to defend
                [pred0CC, gac2TreeCC, gac2RootCC] = treeK(myRow,ccTrain);
                
                % at this point we have build our GAC tree -above executable line- and we can check the
                % instances with the second tree
                [predGac2CC(runCounter,i), predictionZone] = secondGac(gac2TreeCC, gac2RootCC, myRow);
                if predGac2CC(runCounter,i) == -1    % meaning there were not enough instances for gac2 tree
                    i = i-1;
                    continue;
                end
                
                % below part needs to be dynamic
                % now get statistics from the prediction zone
                predictionZoneSizeCC(runCounter,i) = size(predictionZone,1);
                for tmp1 = 1:size(ccIndex,2)
                    eval(['tmpName = [' char(currentDatasetName) '(' num2str(ccIndex(tmp1)) ')];']);
                    tmpDataset = csvread([char(tmpName) '.csv']);
                    tmpDataset = myNormalizer(tmpDataset);
                    [howManySelected] = findMyClassStatistics(predictionZone, tmpDataset);
                    if howManySelected < 0
                        howManySelected
                    end
                    % also define a variable to keep track of how many instances
                    % are selected from each CC center
                    eval([char(tmpName) 'Instances(runCounter,' num2str(i) ') = howManySelected;']);
                end
                
                
                % below are the ones for various k values
                %                 [pred1CC(runCounter,i)] = nnk(myRow,ccTrain,1);
                %                 [pred2CC(runCounter,i)] = nnk(myRow,ccTrain,2);
                %                 [pred4CC(runCounter,i)] = nnk(myRow,ccTrain,4);
                %                 [pred8CC(runCounter,i)] = nnk(myRow,ccTrain,8);
                %                 [pred16CC(runCounter,i)] = nnk(myRow,ccTrain,16);
                %                 [predxCC(runCounter,i)] = nnk(myRow,ccTrain,myBestK);
            end
            runCounter
        end
        
        
        % now that we have done WC and CC experiments for one WC subset
        % we need to derive the win,tie,loss statistics out of them
        [winMmre tieMmre lossMmre medMmre] = winTieLossCalculatorMMRE(actualEffortValues',predGac2WC,predGac2CC);
        [winMdmre tieMdmre lossMdmre medMdmre] = winTieLossCalculatorMdmre(actualEffortValues',predGac2WC,predGac2CC);
        [winPred25 tiePred25 lossPred25 medPred25] = winTieLossCalculatorPred25(actualEffortValues',predGac2WC,predGac2CC);
        [winAr tieAr lossAr medAr] = winTieLossCalculatorAr(actualEffortValues',predGac2WC,predGac2CC);
        
        % now that we are done with the derivation of all the stuff about
        % the subset, we can start writing these things into a file
        resultsFile = fopen('resultsFile.txt','a' );
        winTieLossFileAll = fopen('winTieLossFileAll.txt','a' );
        winTieLossFileMMRE = fopen('winTieLossFileMMRE.txt','a' );
        winTieLossFileMdMRE = fopen('winTieLossFileMdMRE.txt','a' );
        winTieLossFilePred25 = fopen('winTieLossFilePred25.txt','a' );
        winTieLossFileMAR = fopen('winTieLossFileMAR.txt','a' );
        
        fprintf(resultsFile, ['\\textbf{S' num2str(subsetCounter) ': }' char(wcFileName) ' (' num2str(size(wcDataset,1)) ') & ']);
        fprintf(winTieLossFileAll, [char(wcFileName) ' & ']);
        fprintf(winTieLossFileMMRE, [char(wcFileName) ' & ']);
        fprintf(winTieLossFileMdMRE, [char(wcFileName) ' & ']);
        fprintf(winTieLossFilePred25, [char(wcFileName) ' & ']);
        fprintf(winTieLossFileMAR, [char(wcFileName) ' & ']);
       
        
        %%%%% now write the avg. instance sizes of GAC and subsets
        fprintf(resultsFile, [ num2str(mean(mean(predictionZoneSizeCC')),'%6.1f') ]);
        for tmp1 = 1:size(ccIndex,2)
            eval(['tmpName = [' char(currentDatasetName) '(' num2str(ccIndex(tmp1)) ')];']);
            eval(['avgSubsetSize = mean(mean(' char(tmpName) 'Instances''));']);
            myTmpPercentage = (avgSubsetSize*100)/ccDatasetSizes(tmp1);
            fprintf(resultsFile, ' & ');
            if subsetCounter == tmp1
                fprintf(resultsFile, [' {\\G} ' num2str(avgSubsetSize,'%6.1f')]);
            else
                fprintf(resultsFile, [num2str(avgSubsetSize,'%6.1f')]);
            end
            fprintf(resultsFile, [ ' (' num2str(myTmpPercentage,'%6.1f') 'asdf\\%)' ]);
        end
        fprintf(resultsFile, '\\\\ \n');
        
        % write win tie loss values of within company w.r.t. MAR
        fprintf(winTieLossFileMAR, [num2str(winAr(1)) '&' num2str(tieAr(1)) '&' num2str(lossAr(1)) '&   ']);
        fprintf(winTieLossFileMAR, [num2str(medAr(1),'%6.1f') '&' num2str(medAr(2),'%6.1f')]);
        fprintf(winTieLossFileMAR, ' \\\\ \n');
        % write win tie loss values of within company w.r.t. MMRE
        fprintf(winTieLossFileMMRE, [num2str(winMmre(1)) '&' num2str(tieMmre(1)) '&' num2str(lossMmre(1)) '&   ']);
        fprintf(winTieLossFileMMRE, [num2str(medMmre(1),'%6.1f') '&' num2str(medMmre(2),'%6.1f')]);
        fprintf(winTieLossFileMMRE, ' \\\\ \n');
        % write win tie loss values of within company w.r.t. MdMRE
        fprintf(winTieLossFileMdMRE, [num2str(winMdmre(1)) '&' num2str(tieMdmre(1)) '&' num2str(lossMdmre(1)) '&   ']);
        fprintf(winTieLossFileMdMRE, [num2str(medMmre(1),'%6.1f') '&' num2str(medMdmre(2),'%6.1f')]);
        fprintf(winTieLossFileMdMRE, ' \\\\ \n');
        % write win tie loss values of within company w.r.t. Pred25
        fprintf(winTieLossFilePred25, [num2str(winPred25(1)) '&' num2str(tiePred25(1)) '&' num2str(lossPred25(1)) '&   ']);
        fprintf(winTieLossFilePred25, [num2str(medPred25(1),'%6.1f') '&' num2str(medPred25(2),'%6.1f')]);
        fprintf(winTieLossFilePred25, ' \\\\ \n');
        
        % now write all above statistics into winTieLossAll file
        fprintf(winTieLossFileAll, [num2str(winAr(1)) '&' num2str(tieAr(1)) '&' num2str(lossAr(1)) '& ']);
        fprintf(winTieLossFileAll, [num2str(medAr(1),'%6.1e') '&' num2str(medAr(2),'%6.1e') '&   ']);
        fprintf(winTieLossFileAll, [num2str(winMmre(1)) '&' num2str(tieMmre(1)) '&' num2str(lossMmre(1)) '& ']);
        fprintf(winTieLossFileAll, [num2str(medMmre(1),'%6.1f') '&' num2str(medMmre(2),'%6.1f') '&   ']);
        fprintf(winTieLossFileAll, [num2str(winMdmre(1)) '&' num2str(tieMdmre(1)) '&' num2str(lossMdmre(1)) '& ']);
        fprintf(winTieLossFileAll, [num2str(medMdmre(1),'%6.1f') '&' num2str(medMdmre(2),'%6.1f') '&   ']);
        fprintf(winTieLossFileAll, [num2str(winPred25(1)) '&' num2str(tiePred25(1)) '&' num2str(lossPred25(1)) '& ']);
        fprintf(winTieLossFileAll, [num2str(medPred25(1),'%6.1f') '&' num2str(medPred25(2),'%6.1f')]);
        fprintf(winTieLossFileAll, ' \\\\ \n');
    end
    % place a straight line at the end of each dataset
    fprintf(resultsFile, ' \\hline \n');
    fprintf(winTieLossFileAll, ' \\hline \n');
    fprintf(winTieLossFileMAR, ' \\hline \n');
    fprintf(winTieLossFileMMRE, ' \\hline \n');
    fprintf(winTieLossFileMdMRE, ' \\hline \n');
    fprintf(winTieLossFilePred25, ' \\hline \n');
    % finally close the file
    fclose(resultsFile);
    fclose(winTieLossFileAll);
    fclose(winTieLossFileMMRE);
    fclose(winTieLossFileMdMRE);
    fclose(winTieLossFilePred25);
    fclose(winTieLossFileMAR);
end