[Classification_accuracy,p1,m1,classes]=simclass(datalearn, datatest,v,c, measure, p, m,pl)

function [Classification_accuracy,p1,m1,classes]=simclass(datalearn, datatest,v,c, measure, p, m,pl)
%Inputs:
% datalearn: put your data file in matrix form, rows indicates samples and
% columns features of the data.
% datatest: put your data file in matrix form, rows indicates samples and
% columns features of the data.
% v:   how many columns of features you are using e.g. [1:4] means you will
% use first four columns as features of your data.
% c: column where you class labels are. They should be in numerical form.
% measure: Which quantifier in OWA operators you want to use. 
% 1=Basic RIM quantifier
% 2=Quadratic quantifier
% 3=Exponential quantifier
% 4=Trigonometric quantifier
% 5=O'Hagans method.
% p: parameter in generalized Lukasiewics similarity, can be studied as a range of
% parameter values and then given as a vector i.e. p=[0.1:0.1:5].
% alpha: alpha value in OWA operators.
% Can be given as vector i.e. alpha=[0.1:0.1:5].
% pl: do you want to plot how the parameter changes in p and alpha changes the
% mean classification accuracies and variances.
%
%OUTPUTS:
% Mean_accuracy: Mean classification accuracy with best parameter values in
% p and m:
% p1 and m1: best parameter values w.r.t. mean classification accuracy.
% classes: Class information. To which class the sample was classified (in testing set data, datatest) 

sv_name = 'results1'; % mat file where results are stored ('' = "no storing")

[datatest, lc1, cs1] = init_data(datatest,v,c); 
[datalearn, lc2, cs2] = init_data(datalearn,v,c); 

w_opt = 0; % Do we use weight optimization. Not implemented in this version.

%Initializations
N=1;
rn=1;
fitness = zeros(1,N);
fitness_id = zeros(1,N);
fitness_dif = zeros(1,N);
Means = zeros(length(p),length(m),length(rn));
Vars = zeros(length(p),length(m),length(rn)); 
Maxsf = zeros(length(p),length(m),length(rn));
Minsf = zeros(length(p),length(m),length(rn));

Means_fit_dif = zeros(length(p),length(m),length(rn));
Vars_fit_dir = zeros(length(p),length(m),length(rn)); 
Ideal_var = zeros(length(p),length(m), length(lc1),length(rn));

for n = 1 : 1
    rn_ideal = ones(1,length(lc1)); 
    for j = 1:length(m)    
        for i = 1:length(p)  
            y = [p(i), m(j),measure]; % p and m values and similarity measure
        
            for k = 1 : 1 
                ideals(:,:,k) = idealvectors(datalearn, y); % idealvectors         
                if w_opt == 0  
                  [fitness(k), class, Simil] = calcfit(datatest, ideals(:,:,k), y);
                end
            end
            classinfo(i,j,1:length(class))=class;
            Means(i,j,n) = mean(fitness);
            Vars(i,j,n) = var(fitness);
            Maxsf(i,j,n) = max(fitness);
            Minsf(i,j,n) = min(fitness);
            fitness=[];
        end    
    end
    tmp=max(max(Means));
    [p1,m1]=find(tmp==Means);
    Classification_accuracy=Means(p1(1),m1(1));
    classes=classinfo(p1(1),m1(1),:);
    if pl==1
        [X,Y] = meshgrid(m,p);
        figure
        surfc(X,Y,Means(:,:,n))
        title('Classification accuracies')
        xlabel('alpha-values')
        ylabel('p-values')
        zlabel('Classification accuracy')
    end
    clear Y
end
if length(sv_name) ~= 0
    save(sv_name)
end