%% Empty the environment variable
close all;
clear;
clc;
format compact;

%%  Load data
load data1 c1      % mosaic-leaf
load data2 c2      % semi-mosaic-leaf
load data3 c3      % round-leaf

data(1:241,:)=c1(1:241,:);
data(242:530,:)=c2(1:289,:);
data(531:801,:)=c3(1:271,:);

wine=data(:,1:19);
wine_labels=data(:,20);

% Training group and testing group 
train_wine = [wine(1:121,:);wine(242:386,:);wine(531:666,:)];
train_wine_labels = [wine_labels(1:121);wine_labels(242:386);wine_labels(531:666)];

test_wine = [wine(122:241,:);wine(387:530,:);wine(667:801,:)];
test_wine_labels = [wine_labels(122:241);wine_labels(387:530);wine_labels(667:801)];

%% Data preprocessing
% The normalization algorithm
[mtrain,ntrain] = size(train_wine);
[mtest,ntest] = size(test_wine);
dataset = [train_wine;test_wine];
[dataset_scale,ps] = mapminmax(dataset',0,1);
dataset_scale = dataset_scale';
train_wine = dataset_scale(1:mtrain,:);
test_wine = dataset_scale( (mtrain+1):(mtrain+mtest),: );


%% The genetic algorithm (GA) algorithm to select the best c&g
%{
ga_option.maxgen = 200;
ga_option.sizepop = 20; 
ga_option.cbound = [0,100];
ga_option.gbound = [0,100];
ga_option.v = 5;
ga_option.ggap = 0.9;

[bestacc,bestc,bestg] = gaSVMcgForClass(train_wine_labels,train_wine,ga_option);

disp('best c&g');
str = sprintf( 'Best Cross Validation Accuracy = %g%% Best c = %g Best g = %g',bestacc,bestc,bestg);
disp(str);
cmd = ['-c ',num2str(bestc),' -g ',num2str(bestg),'-t',1];

%%  SVM model training
model = svmtrain(train_wine_labels,train_wine,cmd);
%}

%%  model training
model = svmtrain(train_wine_labels, train_wine, '-c 2 -g 1.5 -t 1');

%%  prediction
[predict_label, accuracy] = svmpredict(test_wine_labels, test_wine, model);

%%  Result analysis
figure;
hold on;
plot(test_wine_labels,'o');
plot(predict_label,'r*');
xlabel('training group','FontSize',12);
ylabel('label','FontSize',12);
legend('The classification for actual testing group.','The classification for predicted testing group.');
grid on;

%% 10 times of cross validation
%{
  dataNew=data(:,1:end-1);
  target=data(:,20);
  [M,N]=size(dataNew);
  indices=crossvalind('Kfold',dataNew(1:M,N),10);
  for k=1:10
     test=(indices==k);
     train=~test;
     train_data=dataNew(train,:);
     train_target=target(train,:);
     test_data=dataNew(test,:);
     test_target=target(test,:);
     [mtrain,ntrain] = size(train_data);
     [mtest,ntest] = size(test_data);
     dataset = [train_data;test_data];
     [dataset_scale,ps] = mapminmax(dataset',0,1);
     dataset_scale = dataset_scale';
     train_data = dataset_scale(1:mtrain,:);
     test_data = dataset_scale( (mtrain+1):(mtrain+mtest),: );

     model = svmtrain(train_target, train_data, '-c 2 -g 1.5 -t 1');
     [predict_label, accuracy] = svmpredict(test_target, test_data, model);
     total_index(k)=accuracy(1,1);
  end
 Average_Cross_Validation_Accuracy=sum(total_index(:))/10
 Best_Cross_Validation_Accuracy=max(total_index(:))
%}