求MATLAB代码：要求对给定数据集按照 4:1 比 例随机分为训练集和测试集（注意按照每个类别均匀分组）。在训练集上使用留 一法交叉验证，选出最优参数（包括平衡参数和核参数），使用径向基核（高斯 核），核半径参数与平衡参数取值范围均为[0.001, 0.01, 0.1, 1, 10, 100, 1000]。使 用最优参数和全部训练集，对测试集中样本进行预测，并统计准确率。岭回归的 求解可使用闭式解或 Majorization-Minimization 方法。

以下是MATLAB代码：

% load data
load data.mat

% split data into training and testing sets
[train_data, test_data] = split_data(data, 0.8);

% perform leave-one-out cross validation on training set to find optimal parameters
balance_params = [0.001, 0.01, 0.1, 1, 10, 100, 1000];
kernel_params = [0.001, 0.01, 0.1, 1, 10, 100, 1000];
num_folds = size(train_data, 1);
best_accuracy = 0;
best_balance_param = 0;
best_kernel_param = 0;
for balance_param = balance_params
    for kernel_param = kernel_params
        accuracy = 0;
        for i = 1:num_folds
            % leave-one-out cross validation
            validation_data = train_data(i, :);
            training_data = train_data([1:i-1, i+1:end], :);

            % train SVM
            svm = fitcsvm(training_data(:, 1:end-1), training_data(:, end), ...
                          'KernelFunction', 'rbf', 'KernelScale', kernel_param, ...
                          'BoxConstraint', balance_param);

            % test on validation set
            predicted_labels = predict(svm, validation_data(:, 1:end-1));
            actual_labels = validation_data(:, end);
            accuracy = accuracy + sum(predicted_labels == actual_labels) / length(actual_labels);
        end
        accuracy = accuracy / num_folds;
        if accuracy > best_accuracy
            best_accuracy = accuracy;
            best_balance_param = balance_param;
            best_kernel_param = kernel_param;
        end
    end
end

% train SVM on entire training set with best parameters
svm = fitcsvm(train_data(:, 1:end-1), train_data(:, end), ...
              'KernelFunction', 'rbf', 'KernelScale', best_kernel_param, ...
              'BoxConstraint', best_balance_param);

% test on testing set and compute accuracy
predicted_labels = predict(svm, test_data(:, 1:end-1));
actual_labels = test_data(:, end);
accuracy = sum(predicted_labels == actual_labels) / length(actual_labels);
fprintf('Accuracy: %.2f%%\n', accuracy * 100);

function [train_data, test_data] = split_data(data, train_ratio)
    % split data into training and testing sets
    num_classes = max(data(:, end));
    train_data = [];
    test_data = [];
    for i = 1:num_classes
        class_data = data(data(:, end) == i, :);
        num_train = round(size(class_data, 1) * train_ratio);
        train_indices = randperm(size(class_data, 1), num_train);
        test_indices = setdiff(1:size(class_data, 1), train_indices);
        train_data = [train_data; class_data(train_indices, :)];
        test_data = [test_data; class_data(test_indices, :)];
    end
end

其中 `data` 是数据集，最后一列是标签。`split_data` 函数用于将数据集分成训练集和测试集，按照每个类别均匀分组。`fitcsvm` 函数用于训练 SVM 模型，使用径向基核（高斯核），核半径参数与平衡参数取值范围均为 [0.001, 0.01, 0.1, 1, 10, 100, 1000]。最后输出测试集上的准确率。