看到一些面经中出现了手写K-means,因此自己写了一版,如有错误,欢迎指正。
输入:
raw_data: list()(例如,多个坐标点[[0, 0], [1,1],…])
k: int 簇的个数
mse_limit: float 若更新中心点后的mse和更新前的mse,误差在该值以内,则停止迭代
early_stopping:最大迭代次数
输出:
聚类后的结果:dict(),(例如: {0: [[-0.2, -0.2], [0.1, 0.3], [0.2, 0.2], [0.2, 0]], 1: [[-0.1, 1]],…})
import random
import numpy as np
# 初始化簇心
def get_init_centers(raw_data, k):
return random.sample(raw_data, k)
# 计算距离
def cal_distance(x, y):
return np.linalg.norm(np.array(x) - np.array(y))
# 将各点分配到最近的点, 并计算MSE
def get_cluster_with_mse(raw_data, centers):
distance_sum = 0.0
cluster = {}
for item in raw_data:
flag = -1
min_dis = float('inf')
for i, center_point in enumerate(centers):
dis = cal_distance(item, center_point)
if dis < min_dis:
flag = i
min_dis = dis
if flag not in cluster:
cluster[flag] = []
cluster[flag].append(item)
distance_sum += min_dis**2
return cluster, distance_sum/(len(raw_data)-len(centers))
# 计算各簇的中心点,获取新簇心
def get_new_centers(cluster):
center_points = []
for key in cluster.keys():
center_points.append(np.mean(cluster[key], axis=0)) # axis=0,计算每个维度的平均值
return center_points
# K means主方法
def k_means(raw_data, k, mse_limit, early_stopping):
old_centers = get_init_centers(raw_data, k)
old_cluster, old_mse = get_cluster_with_mse(raw_data, old_centers)
new_mse = 0
count = 0
while np.abs(old_mse - new_mse) > mse_limit and count < early_stopping :
old_mse = new_mse
new_center = get_new_centers(old_cluster)
print(new_center)
new_cluster, new_mse = get_cluster_with_mse(raw_data, new_center)
count += 1
print('mse:',np.abs(new_mse), 'Update times:',count)
return new_cluster