Python implements k-means clustering algorithm_K-Means clustering algorithm
If you want to quickly understand the k-means clustering algorithm_K-Means clustering algorithm, you can refer to the article of this big guy, which is easy to understand: K-means clustering algorithm exam sample questions_K-Means clustering algorithm, author: weixin_39789792
Thanks for this blogger.
This blog is only a note for personal use, if there is any infringement, please contact and delete it.
code details
Note:
The coordinate value of the centroid k cannot be too outrageous . For example, the coordinates of one centroid among the three centroids are far away from all the coordinates in the sample data, which will result in no data in the list sse_k1 (or k2, k3). Make len(sse_kx(x is 1, 2, 3))=0, report error: ZeroDivisionError: division by zero
import pylab as pl
def square_Euclid(x, y):
"""
计算欧几里得距离：
若是两个平面上的点，即（x1，y1），和（x2，y2），那这俩点距离即√( (x1-x2)^2+(y1-y2)^2）；
如果是三维空间中的点，则为√( (x1-x2)^2+(y1-y2)^2+(z1-z2)^2 。
"""
return (x[0] - y[0]) ** 2 + (x[1] - y[1]) ** 2
# 初始化各个点
num_x = []
num_y = []
fl = open('data01.txt') # 点数据存放在data01.txt中
for line in fl.readlines():
curLine = line.strip().split()
num_x.append(float(curLine[0]))
num_y.append(float(curLine[1]))
# 初始化三个质心，经过聚类得到三个分组
k1 = [-1.7, 1]
k2 = [-0.5, 0.5]
k3 = [1, 2]
# 定义三个簇
sse_k1 = []
sse_k2 = []
sse_k3 = []
n = len(num_x)
while True:
sse_k1 = []
sse_k2 = []
sse_k3 = []
for i in range(n):
square_E1 = square_Euclid(k1, [num_x[i], num_y[i]])
square_E2 = square_Euclid(k2, [num_x[i], num_y[i]])
square_E3 = square_Euclid(k3, [num_x[i], num_y[i]])
num_min = min(square_E1, square_E2, square_E3)
# 聚类
if num_min == square_E1:
sse_k1.append(i)
elif num_min == square_E2:
sse_k2.append(i)
elif num_min == square_E3:
sse_k3.append(i)
# 求坐标平均值，以确定新的质心(更新质心坐标)
k1_x = sum([num_x[i] for i in sse_k1]) / len(sse_k1)
k1_y = sum([num_y[i] for i in sse_k1]) / len(sse_k1)
k2_x = sum([num_x[i] for i in sse_k2]) / len(sse_k2)
k2_y = sum([num_y[i] for i in sse_k2]) / len(sse_k2)
k3_x = sum([num_x[i] for i in sse_k3]) / len(sse_k3)
k3_y = sum([num_y[i] for i in sse_k3]) / len(sse_k3)
# 只要有质心的坐标发生改变，则更新质心坐标；若三个质心均无变化，则判定以收敛，聚类结束，退出循环
if k1 != [k1_x, k1_y] or k2 != [k2_x, k2_y] or k3 != [k3_x, k3_y]:
k1 = [k1_x, k1_y]
k2 = [k2_x, k2_y]
k3 = [k3_x, k3_y]
else:
break
# 取聚类后的点坐标
kv1_x = [num_x[i] for i in sse_k1]
kv1_y = [num_y[i] for i in sse_k1]
kv2_x = [num_x[i] for i in sse_k2]
kv2_y = [num_y[i] for i in sse_k2]
kv3_x = [num_x[i] for i in sse_k3]
kv3_y = [num_y[i] for i in sse_k3]
pl.plot(kv1_x, kv1_y, '+')
pl.plot(kv2_x, kv2_y, '.')
pl.plot(kv3_x, kv3_y, '^')
# 坐标系大小依样本数据值范围而定
pl.xlim(-2, 2.5)
pl.ylim(-1, 2.5)
pl.show()
Code reference from: k-means clustering algorithm example, author: weixin_38553466
Result display
Sample data used
Create a data01.txt text document under the file where the .py file is located, and store the following data in
-0.113523119722435 0.305566317246824
-0.0363280337139859 0.110677855003451
0.113494507689856 0.285179031884109
-0.00383816850385252 0.670778674827114
-0.180363593046200 0.394837771823933
0.295367728543231 -0.355182535548782
-0.0296566442720039 0.228722511660635
-0.0930361342677474 0.154377592930645
-0.159633545380951 1.03286272700827
-0.609370592744484 0.0100246598182464
0.164875043625935 0.107920610145671
-0.649661855983650 -0.0264148180075531
-0.0853301136043781 0.194929464533097
-0.0869727732803104 -0.166019322253363
0.267258237150858 0.318664851557507
-0.876324515282669 0.578412914115882
0.290320777421500 0.0269704554131184
-0.164202641138215 -0.0216061750617156
-0.408886348765266 -0.178183406834480
-0.00275690297052195 -0.149757266490323
-0.230897603220972 0.202729565016547
-0.289768125501838 0.299373894453753
0.565273947293806 -0.112025265465832
-0.259434375270518 -0.183038062076565
-0.0622055869197436 0.0178584309105331
-0.281488166956539 -0.282493439656289
0.288003999490542 0.354832178282382
-0.00387861254715821 0.245338598261617
0.0230259610960932 0.304367839506965
0.297069520513791 0.398694925851779
0.213528795047459 -0.0341268311839215
0.248545070529365 -0.182513241920946
-0.674431824833610 0.166219624024427
0.0695478578554150 0.364281641067673
1.52144323033782 1.56356334395462
1.54901744911605 1.44082824131763
1.72628026225810 0.999267392962595
1.34339405843162 1.54435051334828
1.63076888391605 0.822969713727122
1.24625402720513 1.50291563943267
1.49966193305128 1.43962200220279
0.806148334745612 1.59798616598320
1.73765675194197 0.801038214866100
0.688725193167526 1.18560461303177
1.31503430771996 1.25566460922217
1.14051881393761 1.28173391148891
0.883497444350820 1.52712829138676
1.35619761199096 1.47157896393621
-1.41400896645106 1.03490557492282
-1.46921827418174 0.691733912712829
-1.06733046906236 0.945293131396786
-0.789899047908273 1.04583303354796
-0.922550939191143 1.39310184834662
-0.918965347657051 1.44432139992464
-1.03616345036068 1.00166612828372
-1.07715160762591 1.51189230738663
-1.01283275702248 1.46105578965393
-1.48079534886488 1.21031313607727
-0.986518252032434 0.949195019118798
-1.62901492888985 1.53208781532487
-1.05432664597088 1.20897843449092
-1.51323198856773 0.929507861004623
-1.55689740607725 1.32978015955565
-1.39341270591838 1.41557221811715
-1.66195228414799 0.787792125905413
-1.24494832523794 1.84020927229746
-0.898778729417616 0.570410077321060
-1.32885894876685 0.732892764435160
-1.09324537986321 1.63706883409855
-1.19875924554585 1.35282905121539
-0.866788557380931 1.11436578620945
-1.30006378262166 1.25366700524127
-1.15735442373393 1.48126320709162
-0.469640188642725 0.975507100878317
-0.887529056287694 1.54350983044641
-1.54530712190787 1.47051092229069
-0.895890659992745 1.23572220775434
-1.54226615688700 1.31046627190501
-1.24686714416393 1.05116769432966
-1.18900045601094 1.35740905869805
-1.65786095402519 1.03723338930851
-1.37644334323617 1.08018136201292
-1.00479602718570 0.921073237322932
-0.958390570797860 1.56536899409517
-0.761574879786032 1.24176101803965
-1.56925161923031 1.04223861195863
-0.979085655811513 1.46432198217887
-1.14713536403328 1.08846006315455
-0.853944089636229 1.39103904734476