第一节2：DBSCAN算法Python实现和效果展示

Posted 2022-07-04 快乐江湖

tags:

篇首语：本文由小常识网(cha138.com)小编为大家整理，主要介绍了第一节2：DBSCAN算法Python实现和效果展示相关的知识，希望对你有一定的参考价值。

文章目录

五：Python实现
六：效果展示

五：Python实现

代码如下

import random
import numpy as np
import matplotlib.pyplot as plt
import pandas as pd


def dbscan(data_set, eps, min_pts):
    examples_nus = np.shape(data_set)[0]  # 样本数量
    unvisited = [i for i in range(examples_nus)]  # 未被访问的点
    visited = []  # 已被访问的点
    # cluster为输出结果，表示对应元素所属类别
    # 默认是一个长度为examples_nus的值全为-1的列表，-1表示噪声点
    cluster = [-1 for i in range(examples_nus)]

    k = - 1  # 用k标记簇号，如果是-1表示是噪声点
    while len(unvisited) > 0:  # 只要还有没有被访问的点就继续循环
        p = random.choice(unvisited)  # 随机选择一个未被访问对象
        unvisited.remove(p)
        visited.append(p)

        p_nighbor = []  # nighbor为p的eps邻域对象集合，密度直接可达
        for i in range(examples_nus):
            if i != p and np.sqrt(np.sum(np.power(data_set[i, :]-data_set[p, :], 2))) <= eps:  # 计算距离，看是否在邻域内
                p_nighbor.append(i)

        if len(p_nighbor) >= min_pts:  # 如果邻域内对象个数大于min_pts说明是一个核心对象
            k = k+1
            cluster[p] = k  # 表示p它属于k这个簇

            for q in p_nighbor:  # 现在要找该邻域内密度可达
                if q in unvisited:
                    unvisited.remove(q)
                    visited.append(q)

                    # nighbor_pi是pi的eps邻域对象集合
                    q_nighbor = []
                    for j in range(examples_nus):
                        if np.sqrt(np.sum(np.power(data_set[j]-data_set[q], 2))) <= eps and j != q:
                            q_nighbor.append(j)

                    if len(q_nighbor) >= min_pts:  # pi是否是核心对象，通过他的密度直接可达产生p的密度可达
                        cluster[q] = k
                        for t in q_nighbor:
                            if t not in p_nighbor:
                                p_nighbor.append(t)

        else:
            cluster[p] = -1  # 不然就是一个噪声点
    return cluster

六：效果展示

（1）人造数据集

eps=0.08
min_pts=2

import numpy as np
import matplotlib.pyplot as plt
import pandas as pd
import DBSCAN


raw_data = pd.read_csv('./dataset/438-3.csv', header=None)
raw_data.columns = ['X', 'Y']
x_axis = 'X'
y_axis = 'Y'

examples_num = raw_data.shape[0]
train_data = raw_data[[x_axis, y_axis]].values.reshape(examples_num, 2)


min_vals = train_data.min(0)
max_vals = train_data.max(0)
ranges = max_vals - min_vals
normal_data = np.zeros(np.shape(train_data))
nums = train_data.shape[0]
normal_data = train_data - np.tile(min_vals, (nums, 1))
normal_data = normal_data / np.tile(ranges, (nums, 1))

# 参数
eps = 0.08
min_pts = 2
cluster = DBSCAN.dbscan(normal_data, eps, min_pts)


plt.figure(figsize=(12, 5), dpi=80)

# 第一幅图已知标签
plt.subplot(1, 2, 1)

plt.scatter(normal_data[:, 0], normal_data[:, 1], color='black')
plt.title('raw graph')

# 第二幅图聚类结果
plt.subplot(1, 2, 2)

class1_X = []
class1_Y = []
class2_X = []
class2_Y = []
class3_X = []
class3_Y = []
noise_X = []  # 噪声点
noise_Y = []  # 噪声点

for index, value in enumerate(cluster):
    if value == 0:
        class1_X.append(normal_data[index][0])
        class1_Y.append(normal_data[index][1])
    elif value == 1:
        class2_X.append(normal_data[index][0])
        class2_Y.append(normal_data[index][1])
    elif value == 2:
        class3_X.append(normal_data[index][0])
        class3_Y.append(normal_data[index][1])
    elif value == -1:
        noise_X.append(normal_data[index][0])
        noise_Y.append(normal_data[index][1])


plt.scatter(class1_X, class1_Y, c='g', label='class1')
plt.scatter(class2_X, class2_Y, c='r', label='class2')
plt.scatter(class3_X, class3_Y, c='blue', label='class3')
plt.scatter(noise_X, noise_Y, c='black', label='noise')
plt.title('DBSCAN')
plt.legend()
plt.show()