1.matlab函数

% k_means algorithm

% data: 需要分类的数据，列：特征 行：样本

% k: 需要分类数, iter: 迭代次数，默认为样本个数 threshold 为前后两次均值差的平方和的阈值，

%低于该阈值则认为收敛，停止搜索，由于不同情况下阈值差异较大，故默认为0，由增加迭代次数来保证结果的收敛

function [u,myclass] = mykmeans(data, k, iter, threshold)

[row, col] = size(data);

u = data(randint(1,k,[1,row]),:);

u_backup = u;

if iter == 0

    iter = row;

end

diff = zeros(k,col);

for j = 1:iter

myclass = zeros(k,row);

add = zeros(k,col);

for i = 1:row

    for m = 1:k

     diff(m,:) = data(i,:) - u(m,:);

    end

    distance = diag(diff * diff');

    [~, min_num] = min(distance);

    myclass(min_num,i) = 1; % 将属于该类的样本标记为1

    add(min_num,:) = add(min_num,:) + data(i,:);

end

for n = 1:k

    u(n,:) = add(n,:)/length(find(myclass(n,:)==1));

end

  tr = trace((u - u_backup) * (u - u_backup)');

  if(threshold ~= 0 && tr < threshold)

      break;

  end

end

2.C#下混合编程

using System;

using System.Collections.Generic;

using System.Linq;

using System.Text;

using System.IO;

using matPrj;

using MathWorks.MATLAB.NET.Arrays;

using MathWorks.MATLAB.NET.Utility;

namespace mykmeans

{

    class Program

    {

        static void Main(string[] args)

        {

            myMathclass myfun = new myMathclass();

            StreamReader mydata_row = new StreamReader("data.txt");

            string line = null;

            Int16 row = 0;

            while ((line = mydata_row.ReadLine()) != null)

            {

                row++;

            }

            mydata_row.Close();

            double[] data = new double[row * 2];

            StreamReader mydata = new StreamReader("data.txt");

            Int16 i = 0;

            while ((line = mydata.ReadLine()) != null)

            {

                

                foreach (string str in line.Split(','))

                {

                    data[i] = Convert.ToInt16(str);

                    i++;

                }

                

            }

            MWNumericArray m_data = new MWNumericArray(16 , 2, data);

            MWArray[] output = new MWArray[2];

            MWArray[] input = new MWArray[4] { m_data, 2, 0, 0};

            myfun.mykmeans (2, ref output, input);

            MWNumericArray x1 = output[0] as MWNumericArray;

            MWNumericArray x2 = output[1] as MWNumericArray;

            

        }

    }

}

3. 测试数据（分两类）

1,2

2,3

3,4

1,3

2,4

5,8

1,9

2,8

11,34

56,9

110,234

239,300

123,456

232,222

194,254

788,234

4.结果

u =

    8.4000    8.4000

  281.0000  283.3333

myclass =

  Columns 1 through 12

     1     1     1     1     1     1     1     1     1     1     0     0

     0     0     0     0     0     0     0     0     0     0     1     1

  Columns 13 through 16

     0     0     0     0

     1     1     1     1