python课程算法设计并优化

题目

data.txt中是多个点的坐标集合S，每个点的坐标都是一个二维数组，每一行代表一个点的二维坐标，每个二维数组的第一个数字代表横坐标，第二个数字代表纵坐标。现有如下想法，大致内容如下:
设置一个正的常数值d=50，对于集合S中任意两个点，只要它们两个之间的欧式距离不超过d，那么这两点就归为一类。
请尝试自行编写python代码实现上述想法，并显示代码运行时间和最后S被划分成了多少个类。

很显然用DBSCAN算法

例如

实现：

import numpy as np
import random
import time
from sklearn.neighbors import KDTree


def dbscan_tree(data, eps):
    kd = KDTree(data)
    k = 0
    n = data.shape[0]
    label = np.zeros(n)
    cluster = np.zeros(n)
    random_indx = np.random.permutation(n)
    for s in random_indx:
        if label[s] == 0:
            label[s] = 1
            neighbor = list(kd.query_radius(data[s].reshape(1, -1), eps)[0])
            k += 1
            cluster[s] = k
            while len(neighbor) > 0:
                new = random.choice(neighbor)
                neighbor.remove(new)
                if label[new] == 0:
                    label[new] = 1
                    new_neighbor = list(kd.query_radius(data[new].reshape(1, -1), eps)[0])
                    cluster[new] = k
                    neighbor = list(set(neighbor + new_neighbor))
    return k

def dist(a, b):
    m = a.shape[0]
    res = np.zeros((m, m))
    for i in range(m):
        res[i] = np.sum((a[i] - b) ** 2, axis=1)
    return res
def dbscan_low(data, eps):
    D = dist(data, data)
    row = data.shape[0]
    k = 0
    label = np.zeros(row)
    cluster = np.zeros(row)
    random_indx = np.random.permutation(row)
    for s in random_indx:
        if label[s] == 0:
            label[s] = 1
            neighbor = list(np.where(D[s] <= eps)[0])
            k += 1
            cluster[s] = k
            while len(neighbor) > 0:
                new = np.random.choice(neighbor)
                neighbor.remove(new)
                if label[new] == 0:
                    label[new] = 1
                    new_neighbor = list(np.where(D[new] <= eps)[0])
                    cluster[new] = k
                    neighbor = list(set(neighbor + new_neighbor))
    return k


data = np.loadtxt("data.txt")


r1 = 50
r2 = r1*r1
begin = time.time()
print(dbscan_tree(data, r1))    # 通过利用kd-tree对欧拉距离加速了。
end = time.time()
print('kd-tree加速欧氏距离，一共用时', end-begin)

begin = time.time()
print(dbscan_low(data, r2))
end = time.time()
print('正常欧氏距离手算，一共用时', end-begin)