k-最近邻聚类k-Nearest Neighbor

来源：互联网发布：微信一秒语音制作软件编辑：程序博客网时间：2024/06/05 16:30

这里只给出k-最近邻聚类k-Nearest Neighbor的实现过程，算法的原理请自行搜索，代码如下：

k-最近邻聚类算法的主程序：

clc;clear;%读取数据文件,生成点矩阵fileID = fopen('D:\matlabFile\KNN\KNN.txt'); C=textscan(fileID,'%f %f'); fclose(fileID); %显示数组结果 %celldisp(C); %将cell类型转换为矩阵类型,这里只假设原数据为二维属性，且是二维的坐标点 CC_init=cat(2,C{1},C{2});%用来保存初始加载的值 CC=CC_init;  %set key parameter:k k=3; % store all points that satisfy the condition k, construct every two % points to edge edges_all=zeros(2,2,1); %edges count count=1; %4-D matrix to store edges edges=zeros(2,2,k,1); %keep this code safe if size(CC>k+1)     for i=1:size(CC,1)         %extract one point         p=CC(i,:);         CC_temp=CC;         CC_temp(i,:)=[];         % find the nearest k points         points=FindNearNeighbor(k,p,CC_temp);         % merge two points to edge         edge_temp=zeros(2,2,1);         for j=1:size(points,1)             edge=[p;points(j,:)];             edge=sortrows(edge);             edge_temp(:,:,j)=edge;             edges_all(:,:,count)=edge;                count=count+1;         end                    edges(:,:,:,i)=edge_temp;     end end % delete common edge flag=1; while flag     for i=1:size(edges_all,3)-1         %remeber all same edge serial numbers         serial_num=zeros(1,1);         %extract first edge         a=edges_all(:,:,i);         % cmopare first edge with the rest edges         for j=i+1:size(edges_all,3)             %extract second edge             b=edges_all(:,:,j);             if isequal(a,b)                 serial_num=cat(1,serial_num,j);             end         end         %delete all common edges         serial_num(1,:)=[];         if length(serial_num)>=1             edges_all(:,:,serial_num)=[];             break;         end     end     if i==size(edges_all,3)-1         flag=0;     end end %plot all edgesfor i=1:size(edges_all,3)    edge=edges_all(:,:,i);    plot(edge(:,1),edge(:,2));    hold onend

FindNearNeighbor函数的实现如下：

function result=FindNearNeighbor(k,p,points)%store points that satisfy the k condition, first column store distance%between p and one of points,second column store the serial number of the%pointdis_p=zeros(1,2);for i=1:size(points,1)    %extract one point    point=points(i,:);    %compute the distance    dis=sqrt((p(1,1)-point(1,1))^2+(p(1,2)-point(1,2))^2);    dis_p(i,:)=[dis,i];end% sort dis_p set follow first columndis_p=sortrows(dis_p,1);dis_satisfy=dis_p(1:k,:);%return the satisfied pointsresult=points(dis_satisfy(:,2),:);end

KNN.txt实验数据如下，请复制后保存为txt格式

32 3235 4342 3931 3629 4927 3946 3344 4041 3637 3766 6659 6557 6063 5960 6055 6551 6762 6861 6557 64120 120115 124124 119109 125104 109111 117125 130123 109129 110107 117

阅读全文

0 0