% 导入训练集
trainData = readtable('train.xlsx');
X = table2array(trainData(:, 1:end-1)); % 特征
Y = trainData{:, end}; % 类别（多分类）

% 导入测试集 (没有标签的测试集)
testData = readtable('test.xlsx');
X_test = table2array(testData(:, 1:end)); % 测试集特征

% 分割数据集：80% 用于训练，20% 用于验证
cv = cvpartition(size(X, 1), 'HoldOut', 0.2);
X_train = X(training(cv), :); % 80%的训练数据
Y_train = Y(training(cv), :);
X_val = X(test(cv), :); % 20%的验证数据
Y_val = Y(test(cv), :);

% 确保 Y_train 和 Y_val 是数值类型
if iscategorical(Y_train)
    Y_train = double(Y_train); % 转换为数值型
end
if iscategorical(Y_val)
    Y_val = double(Y_val); % 转换为数值型
end

% 确保 X_train 和 X_test 的列数一致
disp(size(X_train)); % 打印 X_train 的大小
disp(size(X_test));  % 打印 X_test 的大小

% 定义模型名称
modelNames = {'Multinomial Logistic Regression', 'Decision Tree', 'Random Forest', ...
              'SVM', 'KNN', 'BP Neural Network', 'GBM', 'AdaBoost'};
models = cell(length(modelNames), 1); % 用于存储模型
predictions = cell(length(modelNames), 1); % 用于存储测试集的预测结果

% 循环遍历每个模型
for i = 1:length(modelNames)
    switch modelNames{i}
        case 'Multinomial Logistic Regression'
            % 多分类逻辑回归
            B = mnrfit(X_train, Y_train, 'model', 'nominal');
            predProb = mnrval(B, X_test); % 预测类别概率
            [~, predictions{i}] = max(predProb, [], 2); % 返回概率最大类别
            
            % 在验证集上进行预测
            predProb_val = mnrval(B, X_val); % 在验证集上进行预测
            [~, val_predictions] = max(predProb_val, [], 2);
        
        case 'Decision Tree'
            % 决策树
            model = fitctree(X_train, Y_train);
            predictions{i} = predict(model, X_test);
            val_predictions = predict(model, X_val); % 在验证集上进行预测
        
        case 'Random Forest'
            % 随机森林
            model = fitcensemble(X_train, Y_train, 'Method', 'Bag');
            predictions{i} = predict(model, X_test);
            val_predictions = predict(model, X_val); % 在验证集上进行预测
        
        case 'SVM'
            % SVM 用于多分类
            model = fitcecoc(X_train, Y_train); % 使用 fitcecoc 来处理多分类
            predictions{i} = predict(model, X_test);
            val_predictions = predict(model, X_val); % 在验证集上进行预测
        
        case 'KNN'
            % K-近邻
            model = fitcknn(X_train, Y_train);
            predictions{i} = predict(model, X_test);
            val_predictions = predict(model, X_val); % 在验证集上进行预测
        
        case 'BP Neural Network'
            % BP 神经网络
            model = feedforwardnet(15); % 10为隐藏层神经元个数
            model = train(model, X_train', full(ind2vec(Y_train'))); % ind2vec用于多类
            nnOutput = model(X_test')'; % 输出神经网络预测结果
            [~, predictions{i}] = max(nnOutput, [], 2); % 选择概率最高的类
            nnOutput_val = model(X_val')'; % 在验证集上进行预测
            [~, val_predictions] = max(nnOutput_val, [], 2); % 选择概率最高的类
        
        case 'GBM'
            % 梯度提升（GBM） - 使用适合多分类的 AdaBoostM2
            model = fitcensemble(X_train, Y_train, 'Method', 'AdaBoostM2'); 
            predictions{i} = predict(model, X_test);
            val_predictions = predict(model, X_val); % 在验证集上进行预测


        
        case 'AdaBoost'
            % AdaBoost
            model = fitcensemble(X_train, Y_train, 'Method', 'AdaBoostM2');
            predictions{i} = predict(model, X_test);
            val_predictions = predict(model, X_val); % 在验证集上进行预测
    end
    
    % 输出每个模型的预测结果为 Excel 文件
    results = table(predictions{i}, 'VariableNames', {'Predictions'});
    writetable(results, ['XXX_' modelNames{i} '_trainend.xlsx']);
    
    % 将预测类别标签添加到测试集最后一列
    testResults = testData;
    testResults.Predictions = predictions{i}; % 添加预测结果列
    writetable(testResults, ['XXX_' modelNames{i} '_test_with_predictions.xlsx']);
    
    % 在验证集上评估模型的精度
    figure;
    cm = confusionchart(Y_val, val_predictions); % 使用Y_val进行混淆矩阵计算
    title(['Confusion Matrix - ' modelNames{i} ' (Validation Set)']);
    
    % 计算并显示精度
    accuracy = sum(diag(cm.NormalizedValues)) / sum(cm.NormalizedValues(:));
    fprintf('%s Validation Accuracy: %.2f%%\n', modelNames{i}, accuracy * 100);
    
    % 可视化精度
    figure;
    bar(accuracy);
    title(['Validation Accuracy of ' modelNames{i}]);
    xlabel('Model');
    ylabel('Accuracy');
    xticks(1);
    xticklabels({modelNames{i}});
end