kmeans

clc ; clear all;
pkg load statistics
pkg load nan

% 1.  LOAD IRIS DATASET
load('wine');

% 2.1 REMOVING OUTLIERS
mean_values = mean(wineInputs);
sigma = std(wineInputs);
n = size(wineInputs, 1);
outlier = abs(wineInputs - mean_values(ones(n,1), :)) > 3*sigma(ones(n,1),:);
n_outliers = sum(outlier);
% data contains only one outlier
% we need to remove the outlier from features (meas) and
% also from labels (species)
wineInputs(any(outlier'),:) = [];
wineTargets(any(outlier'),:) = [];

% 2.2 TRANSFORM DATA TO [0-1]
n_features = size(wineInputs,2);
difference = max(wineInputs) - min(wineInputs);
normalized_features = wineInputs;
for i=1:n_features
  normalized_features(:,i) = (wineInputs(:,i) - min(wineInputs(:,i))) ./ difference(i);
end

wineInputs = normalized_features;
n_wines = size(wineInputs,1);
labels = ones(n_wines,1);
for i=1:n_wines
  if (wineTargets(i,1) == 1)
    labels(i) = 1;
  else
    if (wineTargets(i,2) == 1)
      labels(i) = 2;
    else (wineTargets(i,3) == 1)
      labels(i) = 3;
    end
  end
end

data = [wineInputs , labels]

% divide the matrix by their labels
wine1 = data((data(:,14) == 1),:);
wine2 = data((data(:,14) == 2),:);
wine3 = data((data(:,14) == 3),:);



for i=1:n_features
  subplot(4,4,i);

  boxplot({wineInputs(1:58,i),wineInputs(59:121,i),wineInputs(122:168,i)});
  xlim ("manual")
  xlim([0.0 3.0]);
  ylim ("manual")
  ylim([0.0 1.0]);
  title(i)
end


% prepare sets of 80% and 20% of each category
wine1_20     = wine1(1:size(wine1,1)*20/100,:);
wine2_20 = wine2(1:size(wine2,1)*20/100,:);
wine3_20  = wine3(1:size(wine3,1)*20/100,:);
wine1_80     = wine1(1+size(wine1,1)*int32(20/100):size(wine1,1),:);
wine2_80 = wine2(1+size(wine2,1)*int32(20/100):size(wine2,1),:);
wine3_80  = wine3(1+size(wine3,1)*int32(20/100):size(wine3,1),:);

validation_set   = cell2mat([wine1_20(:,1:13);wine2_20(:,1:13);wine3_20(:,1:13)]);
validation_class = [wine1_20(:,14);wine2_20(:,14);wine3_20(:,14)];
training_set     = cell2mat([wine1_80(:,1:13);wine2_80(:,1:13);wine3_80(:,1:13)]);
training_class   = [wine1_80(:,14);wine2_80(:,14);wine3_80(:,14)];