Python 데이터분석 기초 76 - 밀도 기반 클러스터링(DBSCAN)

 

DBSCAN(Density-based spatial clustering of applications with noise)

밀도 기반의 클러스터링은 점이 세밀하게 몰려 있어서 밀도가 높은 부분을 클러스터링 하는 방식이다. 쉽게 설명하면, 어느점을 기준으로 반경 x내에 점이 n개 이상 있으면 하나의 군집으로 인식하는 방식이다.

 

 

# 밀도 기반 클러스터링 : 데이터가 비선형인 경우 일반적인 계층적/비계층적 클러스터링이 불가. 이를 해결하기 위한 방안.
import matplotlib.pylab as plt
from matplotlib import style
import numpy as np
from sklearn.datasets import make_moons
from sklearn.cluster import KMeans, DBSCAN

# 샘플 데이터
x, y = make_moons(n_samples=200, noise=0.05, random_state=0)
print(x)
print('실제 군집 id :', set(y))

# plt.scatter(x[:, 0], x[:, 1])
# plt.show()

# KMeans로 군집 분류
km = KMeans(n_clusters=2, random_state=0)
pred1 = km.fit_predict(x)
print('예측 군집 id :', pred1[:10])
print('예측 군집 id :', set(pred1))

# Kmeans 군집결과 시각화
    
def plotFunc(x, pr):
    plt.scatter(x[pr==0, 0], x[pr==0, 1], s=40, c='blue', marker='o', label='cluster1')
    plt.scatter(x[pr==1, 0], x[pr==1, 1], s=40, c='red', marker='s', label='cluster2')
    plt.scatter(km.cluster_centers_[:,0], km.cluster_centers_[:,1], s=60, c='black', marker='+', label='centroid')
    plt.legend()
    plt.show()

plotFunc(x, pred1) # 완전한 분리가 완 된다.

# DBSCAN으로 군집 분류
ds = DBSCAN(eps=0.2, min_samples=5, metric='euclidean') # eps : 두 샘플간의 최대거리
pred2 = ds.fit_predict(x)

plotFunc(x, pred2)


<console>
[[ 0.81680544  0.5216447 ]
 [ 1.61859642 -0.37982927]
 [-0.02126953  0.27372826]
 [-1.02181041 -0.07543984]
 [ 1.76654633 -0.17069874]
 [ 1.8820287  -0.04238449]
 [ 0.97481551  0.20999374]
 [ 0.88798782 -0.48936735]
 [ 0.89865156  0.36637762]
 [ 1.11638974 -0.53460385]
 [-0.36380036  0.82790185]
 [ 0.24702417 -0.23856676]
 [ 1.81658658 -0.13088387]
 [ 1.2163905  -0.40685761]
 [-0.8236696   0.64235178]
 [ 0.98065583  0.20850838]
 [ 0.54097175  0.88616823]
 [ 0.348031   -0.30101351]
 [ 0.35181497  0.88827765]
 [-0.77708642  0.82253872]
 [ 1.92590813  0.01214146]
 [ 0.86096723 -0.47653216]
 [ 0.19990695  0.99672359]
 [ 1.2895301  -0.37071087]
 [-0.27847636  1.02538452]
 [ 0.24187916 -0.07627812]
 [ 1.84988768 -0.09773674]
 [ 1.88406869  0.0449402 ]
 [ 0.165822   -0.08613126]
 [ 0.13861369  0.89639036]
 [ 0.89087024  0.52265882]
 [-0.22806587  0.84091882]
 [ 0.98279208 -0.46457771]
 [ 0.04237749  0.19457898]
 [ 0.76422612  0.67223332]
 [ 1.91108938  0.21178339]
 [ 0.43608432 -0.23007221]
 [ 0.96186938  0.09923426]
 [-0.84336684  0.52414334]
 [-0.04122466  0.35721873]
 [ 0.55507653 -0.42493298]
 [-0.4388286   0.85940389]
 [ 0.6532646   0.71235382]
 [ 0.10274835  0.06721414]
 [ 1.5486824  -0.34012196]
 [-0.37318371  0.95506411]
 [ 1.01706978  0.19210044]
 [-0.71923685  0.65476676]
 [ 0.16135772 -0.10771978]
 [ 0.86434045 -0.4594568 ]
 [-0.69717533  0.80133734]
 [ 0.32791175 -0.19619019]
 [ 1.98046734  0.03848682]
 [-0.90479784  0.05723938]
 [ 1.04515397 -0.50020349]
 [ 0.7534213   0.65688005]
 [ 0.54968577  0.73635744]
 [ 1.24038086 -0.47577903]
 [ 0.24918868  0.94246199]
 [-0.20756105  0.99290594]
 [ 0.35136403 -0.29065432]
 [-1.01628753  0.16290244]
 [ 1.78137056 -0.1244931 ]
 [ 0.87423825  0.53065346]
 [ 1.09997644 -0.46733763]
 [-1.07022744  0.2365448 ]
 [-0.15869858  1.01497482]
 [ 1.46569247 -0.3808977 ]
 [ 0.03025209  0.97792142]
 [-0.9365943   0.45674926]
 [ 0.66038307 -0.46576222]
 [-0.99144728  0.40662094]
 [ 0.46339847 -0.46605416]
 [-0.132006    0.52447234]
 [ 0.81566997 -0.42821617]
 [-0.94820947  0.37717096]
 [ 0.05300205  0.18597406]
 [ 0.92648634  0.40988975]
 [ 0.60689997  0.78279323]
 [ 0.72961391 -0.37215252]
 [ 1.9796026   0.12425417]
 [-0.02053902  0.97601558]
 [ 0.63818364 -0.49916763]
 [ 2.00639179  0.44597642]
 [ 0.02315539  0.24035667]
 [-0.35883877  1.02716833]
 [ 0.95414653  0.04177433]
 [-0.33921532  0.96308888]
 [ 0.59950492 -0.39774852]
 [ 1.99019644  0.39360049]
 [ 0.33125729  0.9365782 ]
 [ 0.99460422  0.35063363]
 [ 1.98845457  0.2628361 ]
 [-0.67473718  0.76419738]
 [ 2.00751107  0.3651166 ]
 [ 1.78298331 -0.11490401]
 [ 1.73616653 -0.22781554]
 [ 0.40646216 -0.25422904]
 [-1.02505346  0.24337404]
 [ 0.06414296  0.07759793]
 [ 1.30092145 -0.58089757]
 [ 1.97425572  0.30889897]
 [ 0.03228388  1.07937745]
 [ 1.03086156 -0.02389082]
 [-0.90062492  0.30653639]
 [ 0.08068561  0.29131373]
 [-0.98807765  0.1039765 ]
 [-0.47394435  0.96143212]
 [ 1.54651932 -0.35008497]
 [ 0.23332453  0.89648984]
 [-0.58481687  0.80318956]
 [ 0.0374878   1.02322111]
 [-0.01943215  1.07001032]
 [-0.85323667  0.39896937]
 [ 0.92635535  0.37695326]
 [ 1.43250553 -0.50148981]
 [ 0.60622756  0.66229531]
 [ 1.94401554  0.13685573]
 [ 0.57984414 -0.39868907]
 [ 0.74317519  0.50998316]
 [ 0.87116686  0.54105191]
 [-0.71045745  0.57281877]
 [-0.03081568  0.33644614]
 [-0.0298505   0.99553114]
 [-0.06313347  0.42194174]
 [-0.79223214  0.68354165]
 [ 0.92098434  0.04171051]
 [ 0.17794377  0.04536893]
 [ 1.34934828 -0.3941652 ]
 [ 1.98387143  0.50898445]
 [ 1.00104892  0.27158454]
 [-0.5425424   0.76257612]
 [-0.9969011   0.47226403]
 [ 0.23408511 -0.15381658]
 [ 1.21437019 -0.40862022]
 [ 1.60101745 -0.17940652]
 [ 1.15844202 -0.40408591]
 [-1.00922523  0.2161359 ]
 [ 2.01865957  0.50313426]
 [ 0.88839866  0.39017093]
 [ 0.10170896 -0.01206481]
 [-0.01241966  0.47064905]
 [ 0.44566504  0.94595998]
 [-0.3569344   0.98319206]
 [-0.43845037  0.88374167]
 [ 1.01534178  0.06687469]
 [ 0.2310607   0.01153495]
 [ 1.35098772 -0.44520507]
 [ 0.25423421  1.0205525 ]
 [-0.00586456  0.24919627]
 [ 0.4752852  -0.37028432]
 [ 1.68071768 -0.34775296]
 [ 0.84564282  0.45629647]
 [ 0.34218757  0.90613948]
 [ 0.58741368 -0.35078742]
 [-0.17818292  0.96641541]
 [ 1.25865528 -0.4740009 ]
 [ 0.33542814 -0.18023343]
 [ 0.52630774  0.94876068]
 [ 0.6424051   0.77717105]
 [ 0.15770292  0.04709417]
 [ 1.11178863 -0.5065278 ]
 [ 0.60370903  0.83759912]
 [ 1.48247118 -0.32721961]
 [ 0.39793421 -0.36876588]
 [ 1.67240934 -0.09328043]
 [ 0.47551295  0.85547255]
 [ 0.70605116 -0.42241887]
 [ 1.56418943 -0.34860626]
 [ 0.94012854 -0.57508877]
 [ 0.61400301  0.83833823]
 [-1.07139757  0.02669316]
 [-0.91308996  0.52626435]
 [-0.74824469  0.51823742]
 [ 0.14688241  0.0297201 ]
 [ 0.94362014 -0.44829425]
 [ 1.84489829  0.40601924]
 [-0.66827347  0.69085682]
 [-0.7362418   0.59951884]
 [ 0.60146482  0.72551706]
 [ 1.47437703 -0.37541022]
 [-0.88760005  0.50864517]
 [ 1.92892164  0.18201791]
 [ 1.78673422 -0.27470711]
 [ 1.95130228  0.26574549]
 [ 0.33471666  0.98057089]
 [-0.16884749  0.89206411]
 [ 0.77063994 -0.51750338]
 [-0.88700503  0.36696366]
 [-0.62886492  0.79087211]
 [-0.93006783  0.38754885]
 [ 0.42447858  0.93268774]
 [ 0.80861392  0.53599924]
 [ 0.94000928  0.27111431]
 [-0.01609181  0.37369612]
 [-0.53633385  0.86026837]
 [ 1.88281749  0.24435589]
 [ 0.17575161 -0.007231  ]
 [ 0.12423604  1.00790161]
 [ 1.62152568 -0.22328525]]
실제 군집 id : {0, 1}
예측 군집 id : [1 1 0 0 1 1 1 1 1 1]
예측 군집 id : {0, 1}

 

 

KMeans 를 사용하였을 때 시각화(문제있음)

KMeans를 사용하면 완전한 분리가 되지 않는다.

 

 

DBSCAN 을 사용하였을 때 시각화(문제없음)

DBSCAN을 사용하면 완벽한 군집분류가 완성된다.