Matlab实现LDA（完整源码）.zip

% A demo to LDA(Linear Discriminant Analysis) % include a figure to discribe it % Note:the demo data is watermelon_3.0_alpha.csv % Liu Jingchang 2016/3/9 % log: %% get the coefficients of LDA % get data data = csvread('.\watermelon_3.0_alpha.csv',1,1); % get positive and negative class y = data(:,end); positive_class = data(find(y==1),1:end-1); negative_class = data(find(y==0),1:end-1); % get the average vector of each class mu_1 = mean(positive_class); mu_0 = mean(negative_class); % get the between-class scatter matrix S_b = (mu_0-mu_1)' * (mu_0-mu_1); % get the within-class scatter matrix % get the within-class scatter matrix of positive class temp1 = size(positive_class); n = temp1(1); mu_0_repmat = repmat(mu_0,n,1); delta_0 = (positive_class - mu_0_repmat)' * (positive_class - mu_0_repmat); % get the within-class scatter matrix of negative class temp2 = size(positive_class); n = temp2(1); mu_1_repmat = repmat(mu_1,n,1); delta_1 = (positive_class - mu_1_repmat)' * (positive_class - mu_1_repmat); % S_w = delta_0 + delta_1; % get the eigrnvalue of (S_w)^-1 * S_b eig_value = sort(eig(inv(S_w)*S_b),'descend') %% figure to discrobe LDA % plot the scateer in different style to distinguish class plot(positive_class(:,1),positive_class(:,2),'r+',... negative_class(:,1),negative_class(:,2),'y*') hold on % plot the line of LDA t = 0:0.01:1; plot(t,eig_value(1:end-1)'*t)