mpl_simple2

%tic
%clear;
%clc;
%na kanw random ta data kai na apothikeusw ta apotelesmata
%import data
opts = detectImportOptions('shuflenoise.csv','NumHeaderLines',1);
ctg_smoted_noise = readtable('shuflenoise.csv',opts);
ctg_smoted_noise.Properties.VariableNames([1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27]) = {'A','LB','AC','FM','UC','DL','DS','DP','ASTV','MSTV','ALTV','MLTV','Width','Min','Max','Nmax','Nzeros','Mode','Mean','Median','Variance','Tendency','0','1','2','3','status'}


%summary of dataset
summary(ctg_smoted_noise);


% size table (2644 rows,  columns)
[r c] = size(ctg_smoted_noise);

% seperation of depentent and independent variables
Xsmt = table2array(ctg_smoted_noise(:,2:26));
size_Xsmt = size(Xsmt,1)
Ysmt = table2array(ctg_smoted_noise(:,27));
%[Xsmt, mu, stddev] = normalize(Xsmt);


%save all variables
dep_variables = ctg_smoted_noise.Properties.VariableNames;
Xsmt_variables = ctg_smoted_noise(:,2:26);



 
 %define classes
count_y0_y1 = tabulate(Ysmt)% 
 


%seperate data into training and test sets using cv partition to account
split = round(size_Xsmt*0.70); %70/30 train/test
rng(123);

% random num_points 
seq = randperm(size_Xsmt);

% Split X and y into the same proportion of train and test set

Xsmt_Train = Xsmt(seq(1:split),:);


ysmt_Train = Ysmt(seq(1:split));


Xsmt_Test = Xsmt(seq(split+1:end),:);


ysmt_Test = Ysmt(seq(split+1:end),:);





% %% MultiLayer Perceptron training
% 
%   trainFcn = 'trainscg';
%  
% net = patternnet([5 5]);
% net.layers{1}.transferFcn = 'logsig'; % Sigmoid
% net.layers{2}.transferFcn = 'logsig'; % Sigmoid
% net.layers{3}.transferFcn = 'softmax'
% net.divideMode = 'none';     %training all the data (Xtrain)
%   net.performFcn = 'crossentropy';
% %  %Combines adaptive learning rate with momentum training (traingdx)
%  net.trainFcn='traingdx'; 
% net.trainParam.lr = 0.5; %Learning Rate
% net.trainParam.epochs = 500; 
%  
% % %%Training Phase
%  [net,tr] = train(net, Xsmt_Train, ysmt_Train); 
%  view(net)
%  save('randomNeural','net'); %Save the Model
% 
% 
% %% Evaluate model predictions and get confusion matrix
% 
% Y_predict = net(Xsmt_Test)
% predicted = round(Y_predict);%round results
% error = gsubtract(ysmt_Test,predicted); %Error Target - Predicted
% Performance = crossentropy(net,ysmt_Test,predicted)
% targetind = vec2ind(ysmt_Test);
% predictind = vec2ind(Y_predict);

% view(net)
% figure, plotconfusion(ysmt_Test,predicted)
% figure, plotroc(ysmt_Test,predicted)

%%%bayes opt
XX = Xsmt_Train;
YY = ysmt_Train;
%Index and partition for second cvpartitioned dataset
cvmlp = cvpartition(size(YY,1), 'Holdout', 1/3);
% 
var = [optimizableVariable('networkDepth', [1, 3], 'Type', 'integer');
       optimizableVariable('numHiddenNeurons', [1 20], 'Type', 'integer');
       optimizableVariable('lr', [0.001 1], 'Transform', 'log');
       optimizableVariable('momentum', [0.6 0.95]);
       optimizableVariable('trainFcn', {'traingda', 'traingdm', 'traingdx', 'trainscg', 'trainoss'}, 'Type', 'categorical');
       optimizableVariable('transferFcn', {'logsig', 'poslin', 'tansig', 'purelin'}, 'Type', 'categorical')];
  
% % %% train network
%% Optimization Using crossloss (this is a defined function which train the network)
mnfn = @(X)crossloss(XX', YY', cvmlp, X.networkDepth, X.numHiddenNeurons,...
    X.lr, X.momentum, X.trainFcn, X.transferFcn);
res = bayesopt(mnfn, var,'IsObjectiveDeterministic', false,...
       'AcquisitionFunctionName', 'expected-improvement-plus', ... 
       'MaxObjectiveEvaluations', 200);
X = bestPoint(res);
disp('Optimised')