先贴代码,以后解释。
#include<iostream>
#include<vector>
#include<math.h>
#define k 3
using namespace std;
struct Tuple
{
float attr1;
float attr2;
};
float Dist(Tuple t1,Tuple t2)
{
return sqrt((t1.attr1 - t2.attr1)*(t1.attr1 - t2.attr1) + (t1.attr2 - t2.attr2)*(t1.attr2 - t2.attr2));
}
//决定该样本属于哪一个聚类,传入的是聚类的质心(也是一个组,看作x,y)和一个样本,返回的是label;
int clusterofTuple(Tuple means[], Tuple tuple)
{
float distance = Dist(means[0],tuple);
int label = 0;
for (int i = 0; i < k; i++)
{
if (Dist(means[i],tuple)<distance)
{
label = i;
}
}
return i;
}
//获得蔟集的平方误差,用来判断是否还需要继续迭代,传入的是蔟集的质心,以及所有归类好的样本,装着每个蔟集的容器数组,计算该聚类到自己质心的距离,所有距离的加和,返回所有的平方误差
float getVar(Tuple means[], vector<Tuple> cluster[])
{
float var = 0;
for (int i = 0; i < k; i++)
{
vector<Tuple> t = cluster[i];
for (int j = 0; j < t.size(); j++)
{
var += Dist(means[i], t[j]);
}
}
return var;
}
//计算当前蔟集的质心,输入的是一个蔟集的容器,质心的计算就是对于两个属性累加后除以个数求平均,然后返回质心,所以也要初始化一个质心Tuple t
Tuple getMeans(vector<Tuple> cluster)
{
Tuple t;
int num = cluster.size();
float meanX = 0, meanY = 0;
for (int i = 0; i < num; i++)
{
meanX += cluster[i].attr1;
meanY += cluster[i].attr2;
}
t.attr1 = meanX / num;
t.attr2 = meanY / num;
return t;
}
void Kmeans(vector<tuple> tuples)
{ //定义与初始化
//首先是要定义一个放置分好的蔟,那就是容器组咯,一个容器放一个蔟
//然后还要有放三个质心的数组
vector<Tuple> cluster[k];
Tuple means[k];
//首先设置默认的质心,就是每个组分别是所有tuples里面最前面三个;
for(int i = 0; i < k; i++)
{
means[i].attr1 = tuples[i].attr1;
means[i].attr2 = tuples[i].attr2;
}
//第一次计算距离,进行分类,得到第一次的类标,容器的话是直接用push_back放置进去
int label = 0;
for (int i = 0; i < tuples.size(); i++)
{
label = clusterofTuple(means, tuples[i]);
cluster[label].push_back(tuples[i]);
}
//输出刚开始的蔟
for (int i = 0; i < k; i++)
{
cout << "the num of " << i << endl;
vector<Tuple> t = cluster[i];
for (int j = 0; j < t.size(); j++)
{
cout << t[j].attr1 << " " << t[j].attr2 << " " << endl;
}
}
float oldvar = -1;
float newvar = getvar(means,cluster)
//开始大boss了
while (ads(oldvar,newvar)>1)
{
//1先计算新的3个质心
for (int i = 0; i < k; i++)
{
means[i] = getMeans(cluster[i]);
}
//2清空分号蔟的容器,待会才可以根据新的质心重新分配
for (int i = 0; i < k; i++)
{
cluster[i].clear();
}
//3根据新的质心,对于原来传入的所有数据重新分配
for (int i = 0; i < tuples.size(); i++)
{
label = clusterofTuple(means, tuples[i]);
cluster[label].push_back(tuples[i]);
}
//4最后还是输出一下
for (int i = 0; i < k; i++)
{
vector<Tuple> t = cluster[i];
for (int j = 0; j < t.size(); j++)
{
cout << t.attr1 << t.attr2 << endl;
}
}
}
}
int main()
{
Kmeans(tuples);
}