pca_cluster.m

matlab得毕业设计源码 请选择性的使用 内容详细

clear
clc
%pca处理
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
fid = fopen('E:\毕业设计\data\TrainSample.txt','r');
[A,count]=fscanf(fid,'%d %d',[500,2000]);
fclose(fid);

%词语特征-文档矩阵 及其 协方差矩阵
A = A';
R_A = A' * A;

%求最大的特征值
d = eig(R_A);
[V, D] = eig(R_A);
clear R_A;
s = sum(d);
for i=1:500
    if (sum(d(500-i:500))/s > 0.85)
        break;
    end
end
clear s;
global W_KL;
W_KL = V(:,500-i:500)';
clear V D;

%经pca处理后的样本
global P;
P = A*W_KL';
P = P';
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

%k均值聚类
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
k=500;%聚类后的数目
p_cluster(:,1:k) = cluster(P(:,1:1000), k, i+1, 1000);
p_cluster(:,k+1:2*k) = cluster(P(:,1001:2000), k, i+1, 1000);
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

%建立概率神经网络
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
Tc = ones(1,2*k);
Tc(1,k+1:2*k)=1+Tc(1,k+1:2*k);

%训练概率神经网络
t_cluster = ind2vec(Tc);
spread = 1.8;

%创建改进后的概率神经网络的网络结构
net = newpnnex();

%初始化网络
net = my_init(net, p_cluster, t_cluster, i+1, 1000, spread);
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

%竞争学习算法对其训练
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
p_train_ham = P(:,1:1000);
p_train_spam = P(:,1001:2000);

%训练网络
net = my_train(net, p_train_ham, p_train_spam, i+1, 1000);
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

% %建立概率神经网络
% Tc = ones(1,2*k);
% Tc(1,k+1:2*k)=1+Tc(1,k+1:2*k);
% 
% %训练概率神经网络
% T = ind2vec(Tc);
% spread = 2.7;
% net = newpnn(p_cluster, T, spread);

nn_test_with_pca(net, P, W_KL);