clear;
clc;

%% Dataset Paths
% Uncomment the dataset you want to use

% WHU-RS19 Dataset
% trainData = load('Datasets/feat_trn_fc7_prop_ppr_0.9_RS19.mat');
% trainLabels = load('Datasets/trainLab_0.9_RS19.mat');
% testData = load('Datasets/feat_tst_fc7_prop_ppr_0.9_RS19.mat');
% testLabels = load('Datasets/testLab_0.9_RS19.mat');
% trainFeat = trainData.F_trn;
% trainLabel = trainLabels.trainLab;
% testFeat = testData.F_tst;
% testLabel = testLabels.testLab;

% RSSCN7 Dataset
% Data = load('Datasets/RSSCN.mat');
% featAll = Data.featAll;
% labelAll = Data.labelAll;

% UCM Dataset
% load('E:\Uzma\UC merced\UCMerced_LandUse\UC_Merced_extracted_features_and_labels.mat');
% featAll = [featuresTrain; featuresTest];
% labelAll = [labelsTrain; labelsTest];

%% Parameter Settings
N = 10; % Population size for optimization
max_Iter = 100; % Maximum iterations for optimization
trials = 60; % Number of trials

% Initialize result storage
SfList = cell(trials, 1);
vaccList = zeros(trials, 1);
taccList = zeros(trials, 1);
NfList = zeros(trials, 1);
curveList = zeros(trials, max_Iter);
stat = cell(trials, 1);

%% Trials Loop
for i = 1:trials
    fprintf('\n Trial # %g', i);
    
    % Shuffle dataset
    num_samples = size(featAll, 1);
    perm_indices = randperm(num_samples);
    featShuffled = featAll(perm_indices, :);
    labelShuffled = labelAll(perm_indices, :);
    
    % Split data (80% training, 20% testing)
    trainSize = round(0.8 * num_samples);
    trainFeat = featShuffled(1:trainSize, :);
    trainLabel = labelShuffled(1:trainSize, :);
    testFeat = featShuffled(trainSize+1:end, :);
    testLabel = labelShuffled(trainSize+1:end, :);
    
    % Hold-out validation set (20% of training data)
    HO = cvpartition(trainLabel, 'HoldOut', 0.20);
    
    % Feature selection using Fox Optimization Algorithm
    [sFeat, Sf, Nf, curve] = jWOA(trainFeat, trainLabel, N, max_Iter, HO);
    
    % Evaluate validation accuracy
    vAcc = jKNN(sFeat, trainLabel, HO);
    
    % Evaluate test accuracy
    [tAcc, stat{i}] = jKNNTest(sFeat, trainLabel, testFeat(:, Sf), testLabel);
    
    % Store results
    vaccList(i) = vAcc;
    taccList(i) = tAcc;
    SfList{i} = Sf;
    NfList(i) = Nf;
    curveList(i, :) = curve;
    
    fprintf('\n Validation Accuracy: %g %%', vAcc);
    fprintf('\n Test Accuracy: %g %%', tAcc);
    
    % Save trial results
    filename = sprintf('matFiles/Dataset_Name/Trial_%d.mat', i);
    trFeat = trainFeat(:, Sf);
    tsFeat = testFeat(:, Sf);
    save(filename, 'trFeat', 'trainLabel', 'tsFeat', 'testLabel', 'Nf', 'Sf', 'curve', 'vAcc', 'tAcc');
end

%% Final Results Summary
minvAcc = min(vaccList);
maxvAcc = max(vaccList);
avgAcc = mean(vaccList);
mintAcc = min(taccList);
maxtAcc = max(taccList);
avgNfList = mean(NfList);

fprintf('\n\n------------------------------');
fprintf('\n\t\tFinal Results');
fprintf('\n------------------------------');
fprintf('\n\n\tNum of Trials: %g ', trials);
fprintf('\n\tMaximum Iterations: %g ', max_Iter);
fprintf('\n\tAvg Selected Features: %g ', avgNfList);
fprintf('\n\tAvg Validation Accuracy: %g %%', avgAcc);
fprintf('\n\tMin Validation Accuracy: %g ', minvAcc);
fprintf('\n\tMax Validation Accuracy: %g ', maxvAcc);
fprintf('\n\tMin Test Accuracy: %g ', mintAcc);
fprintf('\n\tMax Test Accuracy: %g ', maxtAcc);
fprintf('\n\n------------------------------\n');

% Save summary results
save('matFiles/Dataset_Name/Final_results.mat', 'trials', 'max_Iter', 'N', 'vaccList', 'SfList', 'NfList', 'curveList', 'avgAcc', 'minvAcc', 'maxvAcc', 'avgNfList', 'taccList', 'stat');

%% Generate Colors for Plotting
numColors = trials;
colors = rand(numColors, 3);

%% Plot Optimization Curves
figure;
for trial = 1:trials
    plot(1:max_Iter, curveList(trial, :), 'LineWidth', 1, 'Color', colors(trial, :), 'DisplayName', ['Trial ' num2str(trial)]);
    hold on;
end