DBSCAN算法实现---Python

 1 def ecludDist(x,y):
 2     return np.sqrt(sum(np.sqrt(np.array(x) - np.array(y))))
 3 
 4 def euclidean_distance(data):
 5     all_points = []
 6     for i in data:
 7         temp = []
 8         for j in data:
 9             temp.append(ecludDist(i,j))
10         all_points.append(temp)
11     return all_points

 1 def classify(z):   #z为通过欧式距离计算所得的矩阵
 2     pts = []
 3     for row in z:
 4         density = np.sum(z.ix[row] < eps)
 5         pts = 0   
 6         if density > MinPts:
 7             pts = 1      #核心点
 8         elif density > 1:
 9             pts = 2　　　　#边界点
10         else:
11             pts = 0　　　　#离群点
12         pts.append(pts)
13 return pts

 1 def point_type:
 2     cluster = dict()
 3     i = 0
 4     for row in z:
 5         cluster[i] = np.where(z.ix[row] < eps)[0]
 6         i = i+1
 7     for i in range(len(cluster)):
 8         for j in range(len(cluster)):
 9             if len(set(cluster[i]) & set(cluster[j])) > 0 and i!=j:
10                 cluster[i] = cluster[i] | cluster[j]
11                 cluster[j] = []

1 def independent_filed(cluster):
2     j = 0
3     result = dict()
4     for i in range(len(cluster)):
5         if len(cluster[i]) >0:    
6             result[j] = cluster[i]
7             j =j+1
8     return result

 1 def mark(df):    #传进来的是转换后的数据
 2     for i in range(result)：   #result 为一个字典型结构
 3         for j in result[i]:
 4             df.at[j,‘type‘] = i     #新建一列，并给它把i值添加上去
 5 
 6 
 7 plt.scatter(
 8     df[‘one‘],
 9     df[‘two‘],
10     c=df[‘type‘])
11