Dbscan的简单 实现

.net 的机器学习资料真的少啊。这个聚类的算法到是很多。但是实现都是java或者c++的。c#真少（也许是我太孤陋寡闻了）。我找了sbscan 算法，那出来给大家分享，请大家看了给指点下。其中距离的计算实现了 欧式距离和cos的距离。大方法一共10几中吧。但是也 java的多。矩阵算法也是java的，c#的少。请大牛见赐个吧。废话不说了，代码如下：

 

using System;
using System.Collections;

 

namespace DBSCAN
{
 /// <summary>
 /// Cluster data using DBSCAN (Density-Based Spatical Clustering of Application with Noise) methed
 /// See "Data Mining" for further information
 /// </summary>
 public sealed class DBSCAN
 {
  public ArrayList DataPoints = new ArrayList(128);
  private ArrayList DP2DP;
  private int m_Core_Num;
  private int m_MinPts;
  private double m_eps;

 

        //增加数据点
  /// <summary>
  /// Add DataPoint to DBSCAN module to cluster
  /// </summary>
  public void AddDataPoint(DataPoint dp)
  {
   DataPoints.Add(dp);
   m_Core_Num = 0;
   m_MinPts = 0;
   m_eps = 0;
  }

 

        //清除数据
  public void RemoveAllDataPoints()
  {
   DataPoints.Clear();
   DP2DP.Clear();
   m_Core_Num = 0;
   m_MinPts = 0;
   m_eps = 0;
  }
        //数据初始化
  public void ResetAllDataPointsState()
  {
   foreach(DataPoint dp in DataPoints)
   {
    dp.class_id = 0;
    dp.core_tag = false;
    dp.used_tag = false;
   }
  }
        //数据初始化
  public void PrepareDBSCAN_Table()
  {
   int dp_count = DataPoints.Count;
   DP2DP = new ArrayList(dp_count);
   for(int i=0;i<dp_count;i++)
   {
    // SortedList use DBSCANSort so that can support duplicate key
    // dp_count also include the point itself
    DP2DP.Add(new SortedList(new DBSCANSort(), dp_count));
   }
   SortedList sl;
   DataPoint dp;
   for(int i=0;i<dp_count;i++)
   {
    sl=(SortedList)DP2DP[i];
    dp=(DataPoint)DataPoints[i];
    for(int j=0;j<dp_count;j++)
    {
     double distance = dp.Distance((DataPoint)DataPoints[j]);
     sl.Add(distance, DataPoints[j]);
    }
   }
  }
        //构建 核心
        /// <summary>
        /// eps 距离  minpts  聚类数
        /// </summary>
        /// <param name="eps"></param>
        /// <param name="MinPts"></param>
        /// <returns></returns>
  public int BuildCorePoint(double eps, int MinPts)
  {
   ResetAllDataPointsState();
   int core_num = 0;
   SortedList sl;
   DataPoint src_dp, des_dp;
   for(int i=0;i<DataPoints.Count;i++)
   {
    sl=(SortedList)DP2DP[i];
    des_dp=(DataPoint)sl.GetByIndex(MinPts);
    src_dp=(DataPoint)DataPoints[i];
    if(src_dp.Distance(des_dp)<eps)
    {
     src_dp.core_tag=true;
     core_num++;
    }
   }
   if(core_num>0)
   {
    m_Core_Num = core_num;
    m_MinPts = MinPts;
    m_eps = eps;
   }
   return core_num;
  }
        //聚类
  public void DBSCAN_Cluster()
  {
   DataPoint dp;
   int current_class_id = 1;
   for(int i=0;i<DataPoints.Count;i++)
   {
    dp=(DataPoint)DataPoints[i];
    if(dp.used_tag==false && dp.core_tag==true)
    {
     dp.class_id = current_class_id;
     dp.used_tag = true;
     CorePointCluster(i, current_class_id);
     current_class_id++;
    }
   }  
  }

 

 
        //核心点聚类
  private void CorePointCluster(int dp_pos, int core_class_id)
  {
   DataPoint src_dp, des_dp;
   SortedList sl=(SortedList)DP2DP[dp_pos];
   src_dp=(DataPoint)sl.GetByIndex(0);
   int i=1;
   des_dp=(DataPoint)sl.GetByIndex(i);
   while(src_dp.Distance(des_dp)<m_eps)
   {
    if(des_dp.used_tag == false)
    {
     des_dp.class_id = core_class_id;
     des_dp.used_tag = true;
     if(des_dp.core_tag == true)
      CorePointCluster(DataPoints.IndexOf(des_dp),core_class_id);
    }
    i++;
    try
    {
     des_dp=(DataPoint)sl.GetByIndex(i);
    }
    catch( ArgumentOutOfRangeException )
    {
     // To avoid eps is too large that out of index
     return;
    }
   }
  }
 }

 

 /// <summary>
 /// DBSCAN DataPoint
 /// </summary>
 public class DataPoint
 {

 

        //是否中心
  public bool core_tag = false;
  
        //是否噪音（无关的点 ）
        public int class_id = 0; // 0 indicate NOISE
  
        //计算过
        public bool used_tag = false;

 

  public double d1; // dimension x-axis
  public double d2; // dimension y-axis
  // dimension n (n>=3) can be extend by inherient this class
  // and reimplement following two method.

 

  public DataPoint(double x, double y)
  {
   d1=x;
   d2=y;
  }

 

        /// <summary>
        /// 距离 ，可以使用  雅阁比系数等 0-1之间
        /// </summary>
        /// <param name="dp"></param>
        /// <returns></returns>
  public double Distance(DataPoint dp)
  {

 

 

 

            if (this != dp)
            {
                double d1sq = (d1 - dp.d1) * (d1 - dp.d1);
                double d2sq = (d2 - dp.d2) * (d2 - dp.d2);
                return Math.Sqrt(d1sq + d2sq);

 

                //计算平方差
            }
            else
                return 0;//同1个点
  }
        /// <summary>
        /// 距离 ，可以使用  雅阁比系数等 0-1之间
        /// </summary>
        /// <param name="dp"></param>
        /// <returns></returns>
        public double CosineDistance(DataPoint dp)
        {
            if (this != dp)
            {
                double d1sq = (d1 * dp.d1) + (d1 * dp.d1);
                double d2sq = Math.Sqrt((d2 * d2) + (dp.d2 + dp.d2));
                 double  temp= d1sq/d2sq;
                 return temp;
                //计算平方差
            }
            else
                return 0;//同1个点
        }
 }
    /// <summary>
    /// 比较数据点的大小
    /// </summary>
 public class DBSCANSort:IComparer
 {
  public int Compare(object x, object y)
  {
   int iResult;
   if((double)x > (double)y)
    iResult = 1;
   else
    iResult = -1;
   return iResult;
  }
 }
}