import numpy as np
import pandas as pd
# 找出最优簇选择(初始)
def initial_value(n, k):
centroids = []
for i in range(k):
lst = []
for j in range(n):
minJ = min(data[:, j])
maxJ = max(data[:, j])
rangeJ = float(maxJ - minJ)
res = minJ + rangeJ * np.random.rand()
lst.append(res)
centroids.append(lst)
return centroids
# 算每个的距离,，取最小距离的索引
def distance(data, centroids):
res = []
for i in data:
lst = []
for j in centroids:
d = np.sqrt(((i-np.array(j))**2).sum())
lst.append(d)
res.append(lst.index(min(lst)))
return res
# 更新簇中心
def update(res, K):
center = []
for k in range(K):
pf = [data[i] for i,j in enumerate(res) if j == k]
fin = np.array(pf).mean(axis=0)
center.append(fin)
return center
def main(k, data, iters):
m, n = np.shape(data)
centroids = initial_value(n, k) # 初始点
for i in range(iters):
res = distance(data, centroids) # 最小距离索引
new_centroids = update(res, k) # 更新簇
if (np.array(new_centroids) == centroids).all(): # 若更新前后中心相同，跳出循环
break
centroids = new_centroids
return res
if __name__ == '__main__':
data = pd.read_csv('data.csv').values
k, iters = 2, 100
m, n = np.shape(data)
result = main(k, data, iters)
print(result)
