linear kernel

gistfile1.txt

%-----------------------------a)-----------------------------
data = parseFile('train.data');
m = size(data,1);
X = data(:,1:1558);
X = str2double(X);
yTemp = data(:,1559);
y = ones(size(yTemp,1),1);
n = size(X,2);

% Converting nonad -> -1 and ad -> 1
for i= 1:length(yTemp)
    string = yTemp{i};
    if strcmp(string,'nonad.') == 0
        y(i,1) = -1;
    else
        y(i,1) = 1;
    end
end

Q = (y*y').*(X*X');

% cvx optimization
cvx_begin
    variable alfa(m);
    minimize( 0.5*alfa'*Q*alfa + (-1).*sum(alfa));
    subject to
    alfa'*y==0;
    alfa >= zeros(m,1);
    alfa <= ones(m,1);
cvx_end
alpha_cvx = alfa;
%------------------------------------------------------------%
%-----------------------------b)-----------------------------%

% calculation of w
w = zeros(1,n);
for i=1:m
    w = w + alfa(i) * y(i) .* X(i,:);
end

% calculation of b
indYminus = find(y == -1);
indYplus = find(y == 1);
setYMinus = zeros(1,length(indYminus));
setYPlus = zeros(1,length(indYplus));

for i=1:length(indYminus)
  setYMinus(i) = w * X(indYminus(i),:)';
end

for i=1:length(indYplus)
  setYPlus(i) = w * X(indYplus(i),:)';
end

b = -(max(setYMinus) + min(setYPlus))/2;


% Finding the accuracy of the test data
dataTest = parseFile('test.data');
m = size(dataTest,1);
X = dataTest(:,1:1558);
X = str2double(X);
yTemp = dataTest(:,1559);
y = ones(size(yTemp,1),1);
n = size(X,2);
for i = 1:length(yTemp)
    string = yTemp{i};
    if strcmp(string,'nonad.') == 0
        y(i,1) = -1;
    else
        y(i,1) = 1;
    end
end

correct = 0;
for i = 1:m
    prediction = w * X(i,:)' + b;
    if prediction > 0 && y(i) == 1
        correct = correct + 1;
    end
    if prediction < 0 && y(i) == 0
        correct = correct + 1;
    end
end

display('The accuracy is : ');
accuracy = (correct / length(y))*100;
display(accuracy);