算法介绍:
本算法是用来在一堆数据中分离出几类数据,比如一堆钞票,将其颜色、尺寸等数据采集,通过分析数据来区分出100元、50元、20元等。主要方法是假设该数据集是有C个类,此时可以在这堆数据中任选10个数据作为初始聚类中心,计算其他数据到这些点的距离的大小,按最小距离将其归为一类。通过不断的迭代,计算选取新的聚类中心,再计算新的距离,按最小距离划分,直到聚类中心的值不再发生变化。
本次所要分离的数据是wine数据集,已知有一堆酒,这些酒来自三个不同的地方,也已经直到这些酒的特征:口味、色度,含糖量等,共有13种特征,现将这些特征采集为数据,每一条数据代表一瓶酒,来区分这些酒的产地。
下图是酒的数据:共有178行数据,也就是178瓶酒,每一行有13个特征数据(第一个1是矫正用的分类数据)
具体程序:
using System;
using System.Collections.Generic;
using System.IO;
using System.Linq;
using System.Text;
using System.Threading.Tasks;
namespace WineIde
{
class Program
{
static void Main(string[] args)
{
List<string[]> WineData = new List<string[]>();
WineData = ReadCsv(@"C:\Users\user\Desktop\wine.txt");
List<double> FirstClass = new List<double>();
List<double> SecondClass = new List<double>();
List<double> ThirdClass = new List<double>();
double[,] NumData = new double[WineData.Count, 13];
for (int i = 0; i < WineData.Count; i++)
{
for (int j = 1; j < 14; j++)
{
NumData[i, j-1] = Double.Parse(WineData[i][j]);
}
}
//FirstClass.Add(0);
//SecondClass.Add(59);
//ThirdClass.Add(130);
// for (int m = 0; m < 4; m++)
//{
for (int i = 0; i < WineData.Count; i++)
{
double First = 0;
double Second = 0;
double Third = 0;
for (int j = 0; j < 1; j++)
{
First += (NumData[i, j] - NumData[0, j]) * (NumData[i, j] - NumData[0, j]);
Second += (NumData[i, j] - NumData[59, j]) * (NumData[i, j] - NumData[59, j]);
Third += (NumData[i, j] - NumData[130, j]) * (NumData[i, j] - NumData[130, j]);
}
if (First < Second && First < Third)
{
FirstClass.Add(i);
}
if (First > Second && Second < Third)
{
SecondClass.Add(i);
}
if (First > Third && Second >Third)
{
ThirdClass.Add(i);
}
}
for (int A = 0; A < 4; A++)
{
double[,] FirstClass1 = new double[1, 13];
double[,] FirstClass2 = new double[1, 13];
double[,] FirstClass3 = new double[1, 13];
foreach (int i in FirstClass)
{
for (int k = 0; k < 13; k++)
{
FirstClass1[0, k] = (FirstClass1[0, k] + NumData[i, k]);
}
}
for (int i = 0; i < 13; i++)
{
FirstClass1[0, i] = FirstClass1[0, i] / FirstClass.Count;
}
foreach (int i in SecondClass)
{
for (int k = 0; k < 13; k++)
{
FirstClass2[0, k] = (FirstClass2[0, k] + NumData[i, k]);
}
}
for (int i = 0; i < 13; i++)
{
FirstClass2[0, i] = FirstClass2[0, i] / SecondClass.Count;
}
foreach (int i in ThirdClass)
{
for (int k = 0; k < 13; k++)
{
FirstClass3[0, k] = (FirstClass3[0, k] + NumData[i, k]);
}
}
for (int i = 0; i < 13; i++)
{
FirstClass3[0, i] = FirstClass3[0, i] / ThirdClass.Count;
}
FirstClass.Clear();
SecondClass.Clear();
ThirdClass.Clear();
for (int i = 0; i < WineData.Count; i++)
{
double First = 0;
double Second = 0;
double Third = 0;
for (int j = 0; j < 13; j++)
{
First += (NumData[i, j] - FirstClass1[0, j]) * (NumData[i, j] - FirstClass1[0, j]);
Second += (NumData[i, j] - FirstClass2[0, j]) * (NumData[i, j] - FirstClass2[0, j]);
Third += (NumData[i, j] - FirstClass3[0, j]) * (NumData[i, j] - FirstClass3[0, j]);
}
if (First < Second && First < Third)
{
FirstClass.Add(i);
}
if (First > Second && Second < Third)
{
SecondClass.Add(i);
}
if (First > Third && Second > Third)
{
ThirdClass.Add(i);
}
}
}
Console.WriteLine("第一类数据:");
for (int i = 0; i < FirstClass.Count; i++)
{
Console.Write("{0} ", FirstClass[i]);
}
Console.WriteLine();
Console.WriteLine("第二类数据:");
for (int i = 0; i < SecondClass.Count; i++)
{
Console.Write("{0} ", SecondClass[i]);
}
Console.WriteLine();
Console.WriteLine("第三类数据:");
for (int i = 0; i < ThirdClass.Count; i++)
{
Console.Write("{0} ", ThirdClass[i]);
}
int num = FirstClass.Count + SecondClass.Count + ThirdClass.Count;
Console.WriteLine("总数是:{0}", num);
}
public static List<string[]> ReadCsv(string PathName)
{
FileStream fs = new FileStream(PathName, FileMode.Open, FileAccess.Read);
StreamReader ReadFile = new StreamReader(fs, Encoding.Default);
string strRead = "";
List<string[]> lsFile = new List<string[]>();
while (strRead!=null)
{
strRead = ReadFile.ReadLine();
if (strRead!=null&&strRead.Length>0)
{
lsFile.Add(strRead.Split(','));
}
}
ReadFile.Close();
return lsFile;
}
}
}
结果:
准确率还有待提高。