%交叉驗證 CV_of_Logistic.m

X=X_Log;
Y=Y_Log;
%從cell中取出 為了簡潔，實際程序中不用
% x1_train=X{1};
% x1_test=X{2};
% x2_train=X{3};
% x2_test=X{4};
% x3_train=X{5};
% x3_test=X{6};
% x4_train=X{7};
% x4_test=X{8};
% x5_train=X{9};
% x5_test=X{10};
% 
% y1_train=Y{1};
% y1_test=Y{2};
% y2_train=Y{3};
% y2_test=Y{4};
% y3_train=Y{5};
% y3_test=Y{6};
% y4_train=Y{7};
% y4_test=Y{8};
% y5_train=Y{9};
% y5_test=Y{10};

%設定step（由實踐證明step選2即可）
step=2;
CV_Logistic=0;
clear Er;
clear s;

for i=1:2:10    %i分別取1，3，5，7，9
    [beta,rate_Logistic((i+1)/2,1)] = Logistic_solve(X{i},Y{i},step);   %rate_Logistic((i+1)/2,1)訓練集的正確率        
%在訓練集1上得到 beta
%在測試集1上驗證
    [y] = Logistic_class(X{i+1},beta);    
%計算L(y,f^)=abs(y-y')/2誤分點  ？？？？  
    Er = abs(Y{i+1}-y);   
    L_Logistic((i+1)/2) = sum(Er);    %錯判的個數
    s = size(X{i+1});   %測試集的大小
    rate_Logistic((i+1)/2,2) = (1-L_Logistic((i+1)/2)/s(1))*100;  %測試集的正確率
    CV_Logistic = CV_Logistic + L_Logistic((i+1)/2);    
    clear Er;
    clear s;
end
CV_Logistic = CV_Logistic/sall(1);  %預測誤差的交叉驗證估計