机器学习：基于UCI葡萄酒数据集进行葡萄酒分类及产地预测，采用PCA+Kmeans、PCA+LVQ、BP神经网络等算法实现

clear all; wine_data = xlsread('wine.xlsx'); method = 'BP';%PK:PCA & Kmeans PL:PCA & LVQ BP:BP Neural Network' rate = 0.7;%训练集70%，测试集30% N = 3; total_cnt = size(wine_data,1); train_cnt = round(total_cnt*rate); test_cnt = total_cnt - train_cnt; rand_idx = randperm(total_cnt); train_idx = rand_idx(1:train_cnt); test_idx = rand_idx(train_cnt+1:total_cnt); train_data = wine_data(train_idx,2:size(wine_data,2)); train_class = wine_data(train_idx,1); test_data = wine_data(test_idx,2:size(wine_data,2)); test_class = wine_data(test_idx,1); dim = size(wine_data,2)-1; %矩阵z-score标准化 train_SM = zeros(train_cnt,dim); data_mean = mean(train_data); data_std = std(train_data); test_SM = zeros(test_cnt,dim); for j = 1:dim train_SM(:,j) = (train_data(:,j) - data_mean(j)) / data_std(j); test_SM(:,j) = (test_data(:,j) - data_mean(j)) / data_std(j); end %三种方法 switch method case 'PK' [DS, com_num, PV, score] = PCA(train_SM, 0.66,'ShowFigure'); %初始化聚类中心 center_init = zeros(3,com_num); for i = 1:N center_init(i,:) = mean(score(train_class==i,:)); end [center, train_flag, ~, ~] = Kmeans(score, N,center_init,'ShowFigure');%k-means聚类 test_score = test_SM * PV; %测试集的评分 test_flag = findMinIdx(test_score, center); case 'PL' [DS, com_num, PV, score] = PCA(train_SM, 0.66,'ShowFigure'); [pv,train_flag] = LVQ(score, train_class, [], (1:N)', 0.1, 1000,'ShowFigure');%LVQ test_flag = findMinIdx(test_SM*PV, pv); case 'BP' target = (train_class - 1) / 2; net = newff(train_SM',target',10,{'tansig','purelin'},'traingdx'); net.divideFcn = ''; net.trainParam.show = 50; net.trainParam.epochs = 1000; net.trainParam.goal = 0.001; net.trainParam.lr = 0.01; net = train(net,train_SM',target'); out = sim(net,train_SM'); train_flag = round(out*2+1)'; out = sim(net,test_SM'); test_flag = round(out*2+1)'; otherwise fprintf('check method\n'); return; end train_correct_cnt = sum(train_flag(:,1) == train_class(:,1)); train_accuracy = train_correct_cnt / train_cnt %训练集准确率 test_correct_cnt = sum(test_flag(:,1) == test_class(:,1)); test_accuracy = test_correct_cnt / test_cnt %训练集准确率 figure; plot(1:test_cnt,test_class(:,1),'bo'); hold on; plot(1:test_cnt,test_flag(:,1),'b*'); legend('实际测试集分类','预测测试集分类');