k-mean的简单实现 机器学习

一、实验目的

实现Kmeans聚类算法

• 实验主要过程与结果

1、算法描述

输入：iris数据集，提取1，3维作为聚类数据；初始化k=3

输出：聚类结果

1. 随机生成k个初始点作为初试聚类中心，由中心代表各聚类；

a = [5.1, 1.4]
b = [4.9, 1.4]
c = [4.7, 1.3]
k = 3

1. 计算所有点到这k个中心点的距离，并将点归到离其最近的聚类；

for i in range(150):
    d1 = eucliDist(a, X[i, 0:2])
    d2 = eucliDist(b, X[i, 0:2])
    d3 = eucliDist(c, X[i, 0:2])
    x = min(d1, d2, d3)
    if (x == d1): A.append(X[i, 0:2])
    if (x == d2): B.append(X[i, 0:2])
    if (x == d3): C.append(X[i, 0:2])

1. 将各个簇中的数据求平均值，作为新的类中心，重复上一步，直到所有的簇不再改变；

x = np.mean(A, axis=0)
y = np.mean(B, axis=0)
t = np.mean(C, axis=0)
x = np.round(x, 1)
y = np.round(y, 1)
t = np.round(t, 1)
if ((a == x).all() and (b == y).all() and (c == t).all()):
    flag = 0
else:
    a = x
    b = y
    c = t

2、扩展任务

对聚类结果进行可视化

for i in range(len(A)):
    q = A[i][0]
    w = A[i][1]
    plt.scatter(q, w, c='r')
for i in range(len(B)):
    q1 = B[i][0]
    w1 = B[i][1]
    plt.scatter(q1, w1, c='b')
for i in range(len(C)):
    q2 = C[i][0]
    w2 = C[i][1]
    plt.scatter(q2, w2, c='g')

plt.show()

1. 最终所有代码:

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

data = pd.read_csv("iris.csv", header=None)
data__value = data.values
z = data__value[:, 0]
t = data__value[:, 2]
X = np.vstack((z, t)).T  # 矩阵


def eucliDist(A, B):
    return np.sqrt(sum(np.power((A - B), 2)))


a = [5.1, 1.4]
b = [4.9, 1.4]
c = [4.7, 1.3]
k = 3
flag = 1
A = []
B = []
C = []
while flag == 1:
    for i in range(150):
        d1 = eucliDist(a, X[i, 0:2])
        d2 = eucliDist(b, X[i, 0:2])
        d3 = eucliDist(c, X[i, 0:2])
        x = min(d1, d2, d3)
        if (x == d1): A.append(X[i, 0:2])
        if (x == d2): B.append(X[i, 0:2])
        if (x == d3): C.append(X[i, 0:2])
    x = np.mean(A, axis=0)
    y = np.mean(B, axis=0)
    t = np.mean(C, axis=0)
    x = np.round(x, 1)
    y = np.round(y, 1)
    t = np.round(t, 1)
    if ((a == x).all() and (b == y).all() and (c == t).all()):
        flag = 0
    else:
        a = x
        b = y
        c = t

# 可视化
for i in range(len(A)):
    q = A[i][0]
    w = A[i][1]
    plt.scatter(q, w, c='r')
for i in range(len(B)):
    q1 = B[i][0]
    w1 = B[i][1]
    plt.scatter(q1, w1, c='b')
for i in range(len(C)):
    q2 = C[i][0]
    w2 = C[i][1]
    plt.scatter(q2, w2, c='g')

plt.show()