重返未来1999航海记提升准确率matlab攻略代码

clc;clear;

% 1. 定义数据和目标范围（替换为实际数据）

box1 = [-1, -1, 0, 1, 3, 2]; % 盒子1的6个数字

box2 = [1, 1, 2, 2, 3, 3]; % 盒子2的6个数字

box3 = [1, 2, 3, 4, 5, 6]; % 盒子3的6个数字

boxes = {box1, box2, box3}; % 存储所有盒子

target_min = 1;   % 目标范围下限

target_max = 7;  % 目标范围上限

% 2. 生成所有可能的3个盒子选择方案（共3^3=27种，允许重复）

schemes = [];

for i = 1:3

for j = 1:3

for k = 1:3

 schemes = [schemes; i, j, k]; % 每行表示一个方案：[盒子A, 盒子B, 盒子C]

end

end

end

num_schemes = size(schemes, 1);

% 3. 为所有方案生成训练数据（特征+标签）

features = [];

labels = [];

scheme_ids = []; % 记录每个样本属于哪个方案

for s = 1:num_schemes

% 当前方案选择的盒子编号

scheme = schemes(s, :); % 获取当前方案

n = length(scheme); % 关键修复：定义n为当前方案中盒子的数量（此处固定为3）

% 生成该方案下所有数字组合（6×6×6=216种）

grids = cell(1, n); % 现在n已正确定义

for j = 1:n

grids{j} = boxes{scheme(j)};

end

[grid_vars{1:n}] = ndgrid(grids{:}); % 生成网格

combinations = cell2mat(cellfun(@(x) x(:), grid_vars, 'UniformOutput', false));

% 提取特征

sum_feat = sum(combinations, 2);  % 特征1：和

prod_feat = prod(combinations, 2);  % 特征2：乘积

sq_sum_feat = sum(combinations.^2, 2);  % 特征3：平方和

max_feat = max(combinations, [], 2);  % 特征4：最大值

min_feat = min(combinations, [], 2);  % 特征5：最小值

current_features = [combinations, sum_feat, prod_feat, sq_sum_feat, max_feat, min_feat];

% 生成标签（1=有效，0=无效）

current_labels = (sum_feat >= target_min) & (sum_feat <= target_max);

current_labels = current_labels(:);

% 汇总数据

features = [features; current_features];

labels = [labels; current_labels];

scheme_ids = [scheme_ids; repmat(s, size(combinations, 1), 1)];

end

% 4. 训练随机森林模型

rng(1); % 固定随机种子，保证结果可复现

idx = randperm(size(features, 1));

train_idx = idx(1:round(0.8*length(idx)));

test_idx = idx(round(0.8*length(idx))+1:end);

% 训练模型

model = fitcensemble(features(train_idx, :), labels(train_idx), ...

'Method', 'Bag', ...  % 集成方法：装袋

'NumLearningCycles', 50, ... % 50棵决策树

'Learner', 'tree');  % 基学习器：决策树

% 评估模型性能

pred = predict(model, features(test_idx, :));

accuracy = mean(pred == labels(test_idx));

fprintf('模型测试集准确率：%.2f%%\n', accuracy*100);

% 5. 预测每个方案的有效概率，找到最优方案

scheme_probs = zeros(num_schemes, 1);

for s = 1:num_schemes

s_idx = scheme_ids == s;

s_features = features(s_idx, :);

s_pred = predict(model, s_features);

scheme_probs(s) = mean(s_pred);

end

% 6. 输出结果

[max_prob, best_idx] = max(scheme_probs);

best_scheme = schemes(best_idx, :);

fprintf('\n最优方案：选择（盒子%d, 盒子%d, 盒子%d）\n', ...

best_scheme(1), best_scheme(2), best_scheme(3));

fprintf('该方案的有效概率：%.2f%%\n', max_prob*100);

免责声明：本文系网络转载或改编，未找到原创作者，版权归原作者所有。如涉及版权，请联系删