E-Commerce Part Ⅵ: K Means Clustering 응용

K Means Clustering¶

 

 

ⅰ. 모듈 불러오기 & DATA 특성 파악¶

In [1]:

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns

In [2]:

df = pd.read_csv("Data/Mall_Customers.csv", index_col = 0)
df.head()

Out[2]:

	Gender	Age	Annual Income (k$)	Spending Score (1-100)
CustomerID
1	Male	19	15	39
2	Male	21	15	81
3	Female	20	16	6
4	Female	23	16	77
5	Female	31	17	40

In [3]:

df.info()

<class 'pandas.core.frame.DataFrame'>
Int64Index: 200 entries, 1 to 200
Data columns (total 4 columns):
 #   Column                  Non-Null Count  Dtype 
---  ------                  --------------  ----- 
 0   Gender                  200 non-null    object
 1   Age                     200 non-null    int64 
 2   Annual Income (k$)      200 non-null    int64 
 3   Spending Score (1-100)  200 non-null    int64 
dtypes: int64(3), object(1)
memory usage: 7.8+ KB

In [4]:

df.describe()

Out[4]:

	Age	Annual Income (k$)	Spending Score (1-100)
count	200.000000	200.000000	200.000000
mean	38.850000	60.560000	50.200000
std	13.969007	26.264721	25.823522
min	18.000000	15.000000	1.000000
25%	28.750000	41.500000	34.750000
50%	36.000000	61.500000	50.000000
75%	49.000000	78.000000	73.000000
max	70.000000	137.000000	99.000000

 

• Missing Value Processing

In [5]:

df.isna().sum()

Out[5]:

Gender                    0
Age                       0
Annual Income (k$)        0
Spending Score (1-100)    0
dtype: int64

 

• Gender 변수 one-hot encoding

In [6]:

df = pd.get_dummies(df, columns= ['Gender'], drop_first=True)

 

 

ⅱ. K Means Clustering¶

In [7]:

from sklearn.cluster import KMeans

model = KMeans(n_clusters=3)
model.fit(df)

model.labels_

Out[7]:

array([2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
       2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
       2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
       2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
       2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
       2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 1, 0, 1, 0, 1, 0, 1, 0, 1,
       0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1,
       0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1,
       0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1,
       0, 1])

 

※ 현재 DF에는 feature가 여러개 있기 때문에 산점도를 그리기 어려움

 

In [8]:

res_df = df.copy()
res_df['label'] = model.labels_

res_df

Out[8]:

	Age	Annual Income (k$)	Spending Score (1-100)	Gender_Male	label
CustomerID
1	19	15	39	1	2
2	21	15	81	1	2
3	20	16	6	0	2
4	23	16	77	0	2
5	31	17	40	0	2
...	...	...	...	...	...
196	35	120	79	0	1
197	45	126	28	0	0
198	32	126	74	1	1
199	32	137	18	1	0
200	30	137	83	1	1

200 rows × 5 columns

In [9]:

res_df.groupby('label').mean()

Out[9]:

	Age	Annual Income (k$)	Spending Score (1-100)	Gender_Male
label
0	40.394737	87.000000	18.631579	0.526316
1	32.692308	86.538462	82.128205	0.461538
2	40.325203	44.154472	49.829268	0.406504

In [10]:

res_df['label'].value_counts()

Out[10]:

2    123
1     39
0     38
Name: label, dtype: int64

 

 

ⅲ-ⅰ. 최적의 K값 찾기, Elbow Method¶

In [11]:

dist_list = []

for n in range(2,16):
    model = KMeans(n_clusters = n)
    model.fit(df)
    dist_list.append(model.inertia_)
    
dist_list

Out[11]:

[212889.44245524294,
 143391.59236035674,
 104414.67534220174,
 75399.61541401486,
 58348.64136331504,
 51575.27793107792,
 44357.32664902663,
 40649.64102453101,
 37132.84983602549,
 34877.838013837994,
 32140.57637686386,
 30214.964718614712,
 28060.734110615635,
 26841.910772642394]

In [12]:

sns.lineplot(x=list(range(2,16)), y=dist_list, marker='o')

Out[12]:

<AxesSubplot:>

 

 

ⅲ-ⅱ. 최적의 K값 찾기, Silhouette Score¶

• Cluster 간의 거리도 계산에 포함( 각각의 Cluster의 거리가 멀수록 좋은 Clustering )

In [13]:

from sklearn.metrics import silhouette_score

silhouette_score(df, model.labels_)

Out[13]:

0.33719682488062047

In [14]:

sil_list = []

for n in range(2,16):
    model = KMeans(n_clusters = n)
    model.fit(df)
    sil_list.append(silhouette_score(df, model.labels_))
    
sil_list

Out[14]:

[0.29307334005502633,
 0.383798873822341,
 0.4052954330641215,
 0.4440669204743008,
 0.45205475380756527,
 0.4347734443683834,
 0.4333505967993175,
 0.41541695889588964,
 0.373801236698759,
 0.36589323758883546,
 0.36291744943287396,
 0.360112430940247,
 0.35380780876759266,
 0.3403664220303368]

In [15]:

sns.lineplot(x=list(range(2, 16)), y=sil_list, marker='o')

Out[15]:

<AxesSubplot:>

Cluster가 6개일 때 실루엣 스코어가 가장 높으므로, 최적은 K값은 6이라 할 수 있다.

 

 

ⅳ. 최적의 K값으로 다시 Modeling¶

In [16]:

model = KMeans(n_clusters = 6)
model.fit(df)

df['label'] = model.labels_
df.head()

Out[16]:

	Age	Annual Income (k$)	Spending Score (1-100)	Gender_Male	label
CustomerID
1	19	15	39	1	5
2	21	15	81	1	4
3	20	16	6	0	5
4	23	16	77	0	4
5	31	17	40	0	5

In [17]:

df.groupby('label').mean()

Out[17]:

	Age	Annual Income (k$)	Spending Score (1-100)	Gender_Male
label
0	32.692308	86.538462	82.128205	0.461538
1	27.000000	56.657895	49.131579	0.342105
2	41.685714	88.228571	17.285714	0.571429
3	56.155556	53.377778	49.088889	0.444444
4	25.272727	25.727273	79.363636	0.409091
5	44.142857	25.142857	19.523810	0.380952

 

• 시각화를 통해 label 별 feature들의 분포 파악

In [18]:

sns.boxplot(data=df, x='label', y='Age')

Out[18]:

<AxesSubplot:xlabel='label', ylabel='Age'>

In [19]:

sns.boxplot(data=df, x='label', y='Annual Income (k$)')

Out[19]:

<AxesSubplot:xlabel='label', ylabel='Annual Income (k$)'>

In [20]:

sns.boxplot(data=df, x='label', y='Spending Score (1-100)')

Out[20]:

<AxesSubplot:xlabel='label', ylabel='Spending Score (1-100)'>

 

 

ⅴ. 주성분 분석, Principle Component Analysis¶

In [21]:

df.drop('label', axis=1, inplace=True)
df.head()

Out[21]:

	Age	Annual Income (k$)	Spending Score (1-100)	Gender_Male
CustomerID
1	19	15	39	1
2	21	15	81	1
3	20	16	6	0
4	23	16	77	0
5	31	17	40	0

In [22]:

from sklearn.decomposition import PCA

pca = PCA(n_components = 2)
pca.fit(df)
pca_data = pca.transform(df)

pca_df = pd.DataFrame(pca_data, columns=['PC1', 'PC2'])
pca_df

Out[22]:

	PC1	PC2
0	-31.869945	-33.001252
1	0.764494	-56.842901
2	-57.408276	-13.124961
3	-2.168543	-53.478590
4	-32.174085	-30.388412
...	...	...
195	58.352515	31.017542
196	19.908001	66.446108
197	58.520804	38.346039
198	20.979130	79.376405
199	72.447693	41.811336

200 rows × 2 columns

In [23]:

plt.figure(figsize=(20, 10))
sns.scatterplot(x=pca_df['PC1'], y=pca_df['PC2'], hue=model.labels_, palette='Set2', s=100)

Out[23]:

<AxesSubplot:xlabel='PC1', ylabel='PC2'>