k-means算法cpp实现

k-means过程：

• 从数据集中随机选取k个点作为初始center
• 迭代（直到一定次数或迭代前后无变化）：
  
  • 计算每个向量最近的center，将其归为该类
  • 计算每一类向量的质心，作为新的center

在此使用欧氏距离

测试数据集（68040*32）：CorelFeatures-mld/ColorHistogram.asc

#include <iostream>
#include <math.h>
#include <stdlib.h>
#include <string.h>
#include <string>
#include <vector>
#include <map>
#include <fstream>

using namespace std;

/**
* @brief 计算欧氏距离
* @param a      n维向量
* @param b      n维向量
* @return       两点间的欧氏距离
*/
double calcMeas(vector<double> a, vector<double> b){
    const int n = a.size();
    double p = 0;
    for(int i=0; i<n; i++){
        p += (a[i]-b[i])*(a[i]-b[i]);
    }
    return sqrt(p);
}

const int m = 68040;//行，向量个数
const int n = 32;//列，维数

int k = 10;//类别数
int n_calc = 100;//最大迭代次数

int main(){
    fstream fs("ColorHistogram.asc");
    vector<vector<double> > data;
    for(int i=0; i<m; i++){
        vector<double> line;
        for(int j=0; j<=n; j++){
            double t;
            fs>>t;
            if(j != 0){
                line.push_back(t);
            }
        }
        data.push_back(line);
    }
    fs.close();
    cout<<"read data over."<<endl;

    //随机选取k个向量作为初始中心
    vector<double> center[k];
    for(int i=0; i<k; i++){
        center[i].resize(n);
        int rand_index = rand()%m;
        for(int j=0; j<n; j++){
            center[i][j] = data[rand_index][j];
        }
    }
    cout<<"rand k-center over."<<endl;

    vector<int> belong;//每个向量所属的类别
    belong.resize(m);

    int pre_num_center[k] = {0};//上次迭代后每个类的向量数
    int num_center[k] = {0};//每个类的向量数
    for(int nc=0; nc<n_calc; nc++){
        cout<<"---------------------------------------"<<endl;
        cout<<"calc: "<<nc<<endl;
        //根据欧式距离进行分类
        for(int i=0; i<m; i++){
            belong[i] = 0;
            double length = calcMeas(data[i], center[0]);
            for(int j=1; j<k; j++){
                double meas = calcMeas(data[i], center[j]);
                if(meas < length){
                    length = meas;
                    belong[i] = j;
                }
            }
        }
        cout<<"classify over."<<endl;

        //重新计算每个类的质心
        vector<double> sum_center[k];
        for(int i=0; i<k; i++){
            sum_center[i].resize(n);
        }
        memset(num_center, 0, sizeof(int)*k);
        for(int i=0; i<m; i++){
            num_center[belong[i]]++;
            for(int j=0; j<n; j++){
                sum_center[belong[i]][j] += data[i][j];
            }
        }
        for(int i=0; i<k; i++){
            for(int j=0; j<n; j++){
                center[i][j] = sum_center[i][j]/num_center[i];
            }
        }
        cout<<"recalc center over."<<endl;
        //打印
        cout<<"num of center:"<<endl;
        for(int i=0; i<k; i++){
            cout<<i<<"\t"<<num_center[i]<<endl;
        }
        //是否结束迭代
        bool redo = false;
        for(int i=0; i<k; i++){
            if(pre_num_center[i] != num_center[i]){
                redo = true;
            }
        }
        for(int i=0; i<k; i++){
            pre_num_center[i] = num_center[i];
        }
        if(redo == false) break;
    }

    return 0;
}