% this function finds the closes k neighbors to a row within a dataset
% and returns the median of those k projects' effort values
% and also the mre value


function [knnMedian, mre, kEffort] = nnkForKernel(myRow,train,kValue)
featureSize = size(train,2);

% % firstly normalize the row and data
trainActualCostValues = train(:,size(train,2));
actualRowCost = myRow(:,size(train,2));
myTempToNormalize = [myRow;train];
myTempToNormalize = (mapstd(myTempToNormalize'))';

% get normalized row and replace cost with actual cost
myRow = myTempToNormalize(1,:);
myRow(:,featureSize) = actualRowCost;

% get normalized train data and replace cost with actual cost
myTempToNormalize(1,:) = [];
train = myTempToNormalize;
train(:,featureSize) = trainActualCostValues;

% below array keeps the kNN effort values
kEffort = zeros(kValue,1);

% below array keeps the distance values
kDistances = zeros(kValue,1);

for counter = 1:kValue
    
   % initially let the first instance have the best distance
   bestDist = sqrt(abs(myRow(1,1:(featureSize-1)).^2 - train(1,1:1:(featureSize-1)).^2)); 
   bestIndex = 1; 
   
   for i=2:size(train,1)
       tempDist = sqrt(abs(myRow(1,1:1:(featureSize-1)).^2 - train(i,1:1:(featureSize-1)).^2));
       if tempDist < bestDist
          bestDist = tempDist;
          bestIndex = i;
       end
   end
   
   % save the effort value of the closest neighbor
   kEffort(counter,1) = train(bestIndex,featureSize);
   kDistances(counter,1) = bestDist;
   
   % delete the closest neighbor for the next run (if there is any)
   train(bestIndex,:) = [];
   
end

% return the median of the kNN
knnMedian = median(kEffort);
kDistances = sort(kDistances);
[B, ]

mre = abs(knnMedian - myRow(1,featureSize))/myRow(1,featureSize);

end