Customer Segmentation (Clustering) 🛍️🛒

¶

1. Problem statement
- 1.1. Introduction
- 1.2. Dataset Features
2. Import Libraries and Data
3. Handling Missing Values
4. Feature Engineering
5. Data Analysis and Visualization
6. Outlier Detection
7. Check for Rare Categories
8. Feature Scaling
9. Dimensionality Reduction
10. Clustering
11. Clusters Analysis and Customers Profiling

¶

1. Problem Statement

¶

1.1. Introduction

Customer Personality Analysis is a detailed analysis of a company’s ideal customers. It helps a business to better understand its customers and makes it easier for them to modify products according to the specific needs, behaviors and concerns of different types of customers.

Customer personality analysis helps a business to modify its product based on its target customers from different types of customer segments. For example, instead of spending money to market a new product to every customer in the company’s database, a company can analyze which customer segment is most likely to buy the product and then market the product only on that particular segment.

¶

1.2. Dataset Features

People:

ID: Customer's unique identifier
Year_Birth: Customer's birth year
Education: Customer's education level
Marital_Status: Customer's marital status
Income: Customer's yearly household income
Kidhome: Number of children in customer's household
Teenhome: Number of teenagers in customer's household
Dt_Customer: Date of customer's enrollment with the company
Recency: Number of days since customer's last purchase
Complain: 1 if the customer complained in the last 2 years, 0 otherwise

Products:

MntWines: Amount spent on wine in last 2 years
MntFruits: Amount spent on fruits in last 2 years
MntMeatProducts: Amount spent on meat in last 2 years
MntFishProducts: Amount spent on fish in last 2 years
MntSweetProducts: Amount spent on sweets in last 2 years
MntGoldProds: Amount spent on gold in last 2 years

Promotion:

NumDealsPurchases: Number of purchases made with a discount
AcceptedCmp1: 1 if customer accepted the offer in the 1st campaign, 0 otherwise
AcceptedCmp2: 1 if customer accepted the offer in the 2nd campaign, 0 otherwise
AcceptedCmp3: 1 if customer accepted the offer in the 3rd campaign, 0 otherwise
AcceptedCmp4: 1 if customer accepted the offer in the 4th campaign, 0 otherwise
AcceptedCmp5: 1 if customer accepted the offer in the 5th campaign, 0 otherwise
Response: 1 if customer accepted the offer in the last campaign, 0 otherwise

Place:

NumWebPurchases: Number of purchases made through the company’s website
NumCatalogPurchases: Number of purchases made using a catalogue
NumStorePurchases: Number of purchases made directly in stores
NumWebVisitsMonth: Number of visits to company’s website in the last month

¶

2. Import Libraries and Data

Back to Table of Contents

In [1]:

# handle table-like data and matrices
import pandas as pd
import numpy as np

# visualisation
import seaborn as sns
import matplotlib.pyplot as plt
import missingno as msno
import plotly.express as px
import plotly.graph_objects as go
from plotly.subplots import make_subplots
import plotly.figure_factory as ff

# preprocessing
from sklearn.preprocessing import StandardScaler

# pca
from sklearn.decomposition import PCA

# clustering
from yellowbrick.cluster import KElbowVisualizer
from sklearn.cluster import KMeans, AgglomerativeClustering

# evaluations
from sklearn.metrics import confusion_matrix

# ignore warnings
import warnings
warnings.filterwarnings('ignore')

# to display the total number columns present in the dataset
pd.set_option('display.max_columns', None)

In [2]:

data = pd.read_csv('marketing_campaign.csv', sep="\t")

¶

3. Handling Missing Values

Back to Table of Contents

let's find if we have missing values in the dataset.

In [3]:

data.isnull().sum()

Out[3]:

ID                      0
Year_Birth              0
Education               0
Marital_Status          0
Income                 24
Kidhome                 0
Teenhome                0
Dt_Customer             0
Recency                 0
MntWines                0
MntFruits               0
MntMeatProducts         0
MntFishProducts         0
MntSweetProducts        0
MntGoldProds            0
NumDealsPurchases       0
NumWebPurchases         0
NumCatalogPurchases     0
NumStorePurchases       0
NumWebVisitsMonth       0
AcceptedCmp3            0
AcceptedCmp4            0
AcceptedCmp5            0
AcceptedCmp1            0
AcceptedCmp2            0
Complain                0
Z_CostContact           0
Z_Revenue               0
Response                0
dtype: int64

In [4]:

msno.matrix(data);

In [5]:

data = data.dropna()

In [6]:

data.isnull().sum()

Out[6]:

ID                     0
Year_Birth             0
Education              0
Marital_Status         0
Income                 0
Kidhome                0
Teenhome               0
Dt_Customer            0
Recency                0
MntWines               0
MntFruits              0
MntMeatProducts        0
MntFishProducts        0
MntSweetProducts       0
MntGoldProds           0
NumDealsPurchases      0
NumWebPurchases        0
NumCatalogPurchases    0
NumStorePurchases      0
NumWebVisitsMonth      0
AcceptedCmp3           0
AcceptedCmp4           0
AcceptedCmp5           0
AcceptedCmp1           0
AcceptedCmp2           0
Complain               0
Z_CostContact          0
Z_Revenue              0
Response               0
dtype: int64

let's find if we have duplicate rows.

In [7]:

data.duplicated().sum()

Out[7]:

¶

4. Feature Engineering

Back to Table of Contents

In [8]:

data.info()

<class 'pandas.core.frame.DataFrame'>
Int64Index: 2216 entries, 0 to 2239
Data columns (total 29 columns):
 #   Column               Non-Null Count  Dtype  
---  ------               --------------  -----  
 0   ID                   2216 non-null   int64  
 1   Year_Birth           2216 non-null   int64  
 2   Education            2216 non-null   object 
 3   Marital_Status       2216 non-null   object 
 4   Income               2216 non-null   float64
 5   Kidhome              2216 non-null   int64  
 6   Teenhome             2216 non-null   int64  
 7   Dt_Customer          2216 non-null   object 
 8   Recency              2216 non-null   int64  
 9   MntWines             2216 non-null   int64  
 10  MntFruits            2216 non-null   int64  
 11  MntMeatProducts      2216 non-null   int64  
 12  MntFishProducts      2216 non-null   int64  
 13  MntSweetProducts     2216 non-null   int64  
 14  MntGoldProds         2216 non-null   int64  
 15  NumDealsPurchases    2216 non-null   int64  
 16  NumWebPurchases      2216 non-null   int64  
 17  NumCatalogPurchases  2216 non-null   int64  
 18  NumStorePurchases    2216 non-null   int64  
 19  NumWebVisitsMonth    2216 non-null   int64  
 20  AcceptedCmp3         2216 non-null   int64  
 21  AcceptedCmp4         2216 non-null   int64  
 22  AcceptedCmp5         2216 non-null   int64  
 23  AcceptedCmp1         2216 non-null   int64  
 24  AcceptedCmp2         2216 non-null   int64  
 25  Complain             2216 non-null   int64  
 26  Z_CostContact        2216 non-null   int64  
 27  Z_Revenue            2216 non-null   int64  
 28  Response             2216 non-null   int64  
dtypes: float64(1), int64(25), object(3)
memory usage: 519.4+ KB

Dt_Customer that indicates the date a customer joined the database is not parsed as DateTime

In [9]:

data['Dt_Customer'] = pd.to_datetime(data['Dt_Customer'])

In [10]:

print("The newest customer's enrolment date in the records:", max(data['Dt_Customer']))
print("The oldest customer's enrolment date in the records:", min(data['Dt_Customer']))

The newest customer's enrolment date in the records: 2014-12-06 00:00:00
The oldest customer's enrolment date in the records: 2012-01-08 00:00:00

Extract the "Age" of a customer by the "Year_Birth" indicating the birth year of the respective person.

In [11]:

data['Age'] = 2015 - data['Year_Birth']

Create another feature "Spent" indicating the total amount spent by the customer in various categories over the span of two years.

In [12]:

data['Spent'] = data['MntWines'] + data['MntFruits'] + data['MntMeatProducts'] + data['MntFishProducts'] + data['MntSweetProducts'] + data['MntGoldProds']

Create another feature "Living_With" out of "Marital_Status" to extract the living situation of couples.

In [13]:

data['Living_With'] = data['Marital_Status'].replace({'Married':'Partner', 'Together':'Partner', 'Absurd':'Alone', 'Widow':'Alone', 'YOLO':'Alone', 'Divorced':'Alone', 'Single':'Alone'})

Create a feature "Children" to indicate total children in a household that is, kids and teenagers.

In [14]:

data['Children'] = data['Kidhome'] + data['Teenhome']

To get further clarity of household, Creating feature indicating "Family_Size"

In [15]:

data['Family_Size'] = data['Living_With'].replace({'Alone': 1, 'Partner':2}) + data['Children']

Create a feature "Is_Parent" to indicate parenthood status

In [16]:

data['Is_Parent'] = np.where(data.Children > 0, 1, 0)

Segmenting education levels in three groups

In [17]:

data['Education'] = data['Education'].replace({'Basic':'Undergraduate', '2n Cycle':'Undergraduate', 'Graduation':'Graduate', 'Master':'Postgraduate', 'PhD':'Postgraduate'})

Dropping some of the redundant features

In [18]:

to_drop = ['Marital_Status', 'Dt_Customer', 'Z_CostContact', 'Z_Revenue', 'Year_Birth', 'ID']
data = data.drop(to_drop, axis=1)

In [19]:

data.head(3)

Out[19]:

	Education	Income	Kidhome	Teenhome	Recency	MntWines	MntFruits	MntMeatProducts	MntFishProducts	MntSweetProducts	MntGoldProds	NumDealsPurchases	NumWebPurchases	NumCatalogPurchases	NumStorePurchases	NumWebVisitsMonth	Response	Age	Spent	Living_With	Children	Family_Size	Is_Parent
0	Graduate	58138.0	0	0	58	635	88	546	172	88	88	3	8	10	4	7	1	58	1617	Alone	0	1	0
1	Graduate	46344.0	1	1	38	11	1	6	2	1	6	2	1	1	2	5	0	61	27	Alone	2	3	1
2	Graduate	71613.0	0	0	26	426	49	127	111	21	42	1	8	2	10	4	0	50	776	Partner	0	2	0

¶

5. Data Analysis and Visualization

Back to Table of Contents

In [20]:

data.shape

Out[20]:

(2216, 29)

There are 2216 cutomers and 29 features in the dataset.

In [21]:

data.info()

<class 'pandas.core.frame.DataFrame'>
Int64Index: 2216 entries, 0 to 2239
Data columns (total 29 columns):
 #   Column               Non-Null Count  Dtype  
---  ------               --------------  -----  
 0   Education            2216 non-null   object 
 1   Income               2216 non-null   float64
 2   Kidhome              2216 non-null   int64  
 3   Teenhome             2216 non-null   int64  
 4   Recency              2216 non-null   int64  
 5   MntWines             2216 non-null   int64  
 6   MntFruits            2216 non-null   int64  
 7   MntMeatProducts      2216 non-null   int64  
 8   MntFishProducts      2216 non-null   int64  
 9   MntSweetProducts     2216 non-null   int64  
 10  MntGoldProds         2216 non-null   int64  
 11  NumDealsPurchases    2216 non-null   int64  
 12  NumWebPurchases      2216 non-null   int64  
 13  NumCatalogPurchases  2216 non-null   int64  
 14  NumStorePurchases    2216 non-null   int64  
 15  NumWebVisitsMonth    2216 non-null   int64  
 16  AcceptedCmp3         2216 non-null   int64  
 17  AcceptedCmp4         2216 non-null   int64  
 18  AcceptedCmp5         2216 non-null   int64  
 19  AcceptedCmp1         2216 non-null   int64  
 20  AcceptedCmp2         2216 non-null   int64  
 21  Complain             2216 non-null   int64  
 22  Response             2216 non-null   int64  
 23  Age                  2216 non-null   int64  
 24  Spent                2216 non-null   int64  
 25  Living_With          2216 non-null   object 
 26  Children             2216 non-null   int64  
 27  Family_Size          2216 non-null   int64  
 28  Is_Parent            2216 non-null   int64  
dtypes: float64(1), int64(26), object(2)
memory usage: 519.4+ KB

In [22]:

data.describe()

Out[22]:

	Income	Kidhome	Teenhome	Recency	MntWines	MntFruits	MntMeatProducts	MntFishProducts	MntSweetProducts	MntGoldProds	NumDealsPurchases	NumWebPurchases	NumCatalogPurchases	NumStorePurchases	NumWebVisitsMonth	AcceptedCmp3	AcceptedCmp4	AcceptedCmp5	AcceptedCmp1	AcceptedCmp2	Complain	Response	Age	Spent	Children	Family_Size	Is_Parent
count	2216.000000	2216.000000	2216.000000	2216.000000	2216.000000	2216.000000	2216.000000	2216.000000	2216.000000	2216.000000	2216.000000	2216.000000	2216.000000	2216.000000	2216.000000	2216.000000	2216.000000	2216.000000	2216.000000	2216.000000	2216.000000	2216.000000	2216.000000	2216.000000	2216.000000	2216.000000	2216.000000
mean	52247.251354	0.441787	0.505415	49.012635	305.091606	26.356047	166.995939	37.637635	27.028881	43.965253	2.323556	4.085289	2.671029	5.800993	5.319043	0.073556	0.074007	0.073105	0.064079	0.013538	0.009477	0.150271	46.179603	607.075361	0.947202	2.592509	0.714350
std	25173.076661	0.536896	0.544181	28.948352	337.327920	39.793917	224.283273	54.752082	41.072046	51.815414	1.923716	2.740951	2.926734	3.250785	2.425359	0.261106	0.261842	0.260367	0.244950	0.115588	0.096907	0.357417	11.985554	602.900476	0.749062	0.905722	0.451825
min	1730.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	19.000000	5.000000	0.000000	1.000000	0.000000
25%	35303.000000	0.000000	0.000000	24.000000	24.000000	2.000000	16.000000	3.000000	1.000000	9.000000	1.000000	2.000000	0.000000	3.000000	3.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	38.000000	69.000000	0.000000	2.000000	0.000000
50%	51381.500000	0.000000	0.000000	49.000000	174.500000	8.000000	68.000000	12.000000	8.000000	24.500000	2.000000	4.000000	2.000000	5.000000	6.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	45.000000	396.500000	1.000000	3.000000	1.000000
75%	68522.000000	1.000000	1.000000	74.000000	505.000000	33.000000	232.250000	50.000000	33.000000	56.000000	3.000000	6.000000	4.000000	8.000000	7.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	56.000000	1048.000000	1.000000	3.000000	1.000000
max	666666.000000	2.000000	2.000000	99.000000	1493.000000	199.000000	1725.000000	259.000000	262.000000	321.000000	15.000000	27.000000	28.000000	13.000000	20.000000	1.000000	1.000000	1.000000	1.000000	1.000000	1.000000	1.000000	122.000000	2525.000000	3.000000	5.000000	1.000000

In [23]:

data.describe(include=object).T

Out[23]:

	count	unique	top	freq
Education	2216	3	Graduate	1116
Living_With	2216	2	Partner	1430

In [49]:

sns.pairplot(data , vars=['Spent','Income','Age'] , hue='Children', palette='husl');

In [37]:

plt.figure(figsize=(13,8))
sns.scatterplot(x=data[data['Income']<600000]['Spent'], y=data[data['Income']<600000]['Income'], color='#cc0000');

In [31]:

plt.figure(figsize=(13,8))
sns.scatterplot(x=data['Spent'], y=data['Age']);

In [26]:

plt.figure(figsize=(13,8))
sns.histplot(x=data['Spent'], hue=data['Education']);

In [57]:

data['Education'].value_counts().plot.pie(explode=[0.1,0,0], autopct='%1.1f%%', shadow=True, figsize=(8,8), colors=sns.color_palette('bright'));

¶

6. Outlier Detection

Back to Table of Contents

The presence of outliers in a classification or regression dataset can result in a poor fit and lower predictive modeling performance, therefore we should see there are ouliers in the data.

In [52]:

plt.figure(figsize=(13,8))
sns.distplot(data.Age, color='purple');

In [53]:

plt.figure(figsize=(13,8))
sns.distplot(data.Income, color='Yellow');

In [54]:

plt.figure(figsize=(13,8))
sns.distplot(data.Spent, color='#ff9966');

Another way of visualising outliers is using boxplots and whiskers, which provides the quantiles (box) and inter-quantile range (whiskers), with the outliers sitting outside the error bars (whiskers).

All the dots in the plot below are outliers according to the quantiles + 1.5 IQR rule

In [28]:

fig = make_subplots(rows=1, cols=3)

fig.add_trace(go.Box(y=data['Age'], notched=True, name='Age', marker_color = '#6699ff', 
                     boxmean=True, boxpoints='suspectedoutliers'), 1, 2)

fig.add_trace(go.Box(y=data['Income'], notched=True, name='Income', marker_color = '#ff0066', 
                     boxmean=True, boxpoints='suspectedoutliers'), 1, 1)

fig.add_trace(go.Box(y=data['Spent'], notched=True, name='Spent', marker_color = 'lightseagreen', 
                     boxmean=True, boxpoints='suspectedoutliers'), 1, 3)

fig.update_layout(title_text='<b>Box Plots for Numerical Variables<b>')

fig.show()

In [29]:

data.head(1)

Out[29]:

	Education	Income	Kidhome	Teenhome	Recency	MntWines	MntFruits	MntMeatProducts	MntFishProducts	MntSweetProducts	MntGoldProds	NumDealsPurchases	NumWebPurchases	NumCatalogPurchases	NumStorePurchases	NumWebVisitsMonth	AcceptedCmp3	AcceptedCmp4	AcceptedCmp5	AcceptedCmp1	AcceptedCmp2	Complain	Response	Age	Spent	Living_With	Children	Family_Size	Is_Parent
0	Graduate	58138.0	0	0	58	635	88	546	172	88	88	3	8	10	4	7	0	0	0	0	0	0	1	58	1617	Alone	0	1	0

In [30]:

numerical = ['Income', 'Recency', 'Age', 'Spent']

In [31]:

def detect_outliers(d):
  for i in d:
    Q3, Q1 = np.percentile(data[i], [75 ,25])
    IQR = Q3 - Q1

    ul = Q3+1.5*IQR
    ll = Q1-1.5*IQR

    outliers = data[i][(data[i] > ul) | (data[i] < ll)]
    print(f'*** {i} outlier points***', '\n', outliers, '\n')

In [32]:

detect_outliers(numerical)

*** Income outlier points*** 
 164     157243.0
617     162397.0
655     153924.0
687     160803.0
1300    157733.0
1653    157146.0
2132    156924.0
2233    666666.0
Name: Income, dtype: float64 

*** Recency outlier points*** 
 Series([], Name: Recency, dtype: int64) 

*** Age outlier points*** 
 192    115
239    122
339    116
Name: Age, dtype: int64 

*** Spent outlier points*** 
 1179    2525
1492    2524
1572    2525
Name: Spent, dtype: int64

We will delete some of the outlier points.

In [33]:

data = data[(data['Age']<100)]
data = data[(data['Income']<600000)]

In [34]:

data.shape

Out[34]:

(2212, 29)

¶

7. Check for Rare Categories

Back to Table of Contents

Some categories may appear a lot in the dataset, whereas some other categories appear only in a few number of observations.

Rare values in categorical variables tend to cause over-fitting, particularly in tree based methods.
Rare labels may be present in training set, but not in test set, therefore causing over-fitting to the train set.
Rare labels may appear in the test set, and not in the train set. Thus, the machine learning model will not know how to evaluate it.

In [35]:

categorical = [var for var in data.columns if data[var].dtype=='O']

In [36]:

# check the number of different labels
for var in categorical:
    print(data[var].value_counts() / np.float(len(data)))
    print()
    print()

Graduate         0.504069
Postgraduate     0.382007
Undergraduate    0.113924
Name: Education, dtype: float64


Partner    0.64557
Alone      0.35443
Name: Living_With, dtype: float64

As shown above, there is no rare category in the categorical variables.

¶

7. Categorical Variables Encoding

Back to Table of Contents

In [37]:

categorical

Out[37]:

['Education', 'Living_With']

In [38]:

data['Living_With'].unique()

Out[38]:

array(['Alone', 'Partner'], dtype=object)

Since the education is a ordinal variable, we will encode it with ordinal numbers.

In [39]:

data['Education'] = data['Education'].map({'Undergraduate':0,'Graduate':1, 'Postgraduate':2})

In [40]:

data['Living_With'] = data['Living_With'].map({'Alone':0,'Partner':1})

In [41]:

data.dtypes

Out[41]:

Education                int64
Income                 float64
Kidhome                  int64
Teenhome                 int64
Recency                  int64
MntWines                 int64
MntFruits                int64
MntMeatProducts          int64
MntFishProducts          int64
MntSweetProducts         int64
MntGoldProds             int64
NumDealsPurchases        int64
NumWebPurchases          int64
NumCatalogPurchases      int64
NumStorePurchases        int64
NumWebVisitsMonth        int64
AcceptedCmp3             int64
AcceptedCmp4             int64
AcceptedCmp5             int64
AcceptedCmp1             int64
AcceptedCmp2             int64
Complain                 int64
Response                 int64
Age                      int64
Spent                    int64
Living_With              int64
Children                 int64
Family_Size              int64
Is_Parent                int64
dtype: object

In [42]:

data.head(3)

Out[42]:

	Education	Income	Kidhome	Teenhome	Recency	MntWines	MntFruits	MntMeatProducts	MntFishProducts	MntSweetProducts	MntGoldProds	NumDealsPurchases	NumWebPurchases	NumCatalogPurchases	NumStorePurchases	NumWebVisitsMonth	Response	Age	Spent	Living_With	Children	Family_Size	Is_Parent
0	1	58138.0	0	0	58	635	88	546	172	88	88	3	8	10	4	7	1	58	1617	0	0	1	0
1	1	46344.0	1	1	38	11	1	6	2	1	6	2	1	1	2	5	0	61	27	0	2	3	1
2	1	71613.0	0	0	26	426	49	127	111	21	42	1	8	2	10	4	0	50	776	1	0	2	0

In [43]:

corrmat = data.corr()

plt.figure(figsize=(20,20))  
sns.heatmap(corrmat, annot = True, cmap = 'mako', center = 0)

Out[43]:

<matplotlib.axes._subplots.AxesSubplot at 0x7f6e9dc6b3d0>

¶

8. Feature Scaling

Back to Table of Contents

In this section, numerical features are scaled.

StandardScaler = $\frac{x-\mu}{s}$

In [44]:

data_old = data.copy()

In [45]:

# creating a subset of dataframe by dropping the features on deals accepted and promotions
cols_del = ['AcceptedCmp3', 'AcceptedCmp4', 'AcceptedCmp5', 'AcceptedCmp1','AcceptedCmp2', 'Complain', 'Response']
data = data.drop(cols_del, axis=1)

In [46]:

scaler = StandardScaler()
data = pd.DataFrame(scaler.fit_transform(data), columns = data.columns)

In [47]:

data.head(3)

Out[47]:

	Education	Income	Kidhome	Teenhome	Recency	MntWines	MntFruits	MntMeatProducts	MntFishProducts	MntSweetProducts	MntGoldProds	NumDealsPurchases	NumWebPurchases	NumCatalogPurchases	NumStorePurchases	NumWebVisitsMonth	Age	Spent	Living_With	Children	Family_Size	Is_Parent
0	-0.411675	0.287105	-0.822754	-0.929699	0.310353	0.977660	1.552041	1.690293	2.453472	1.483713	0.852576	0.351030	1.426865	2.503607	-0.555814	0.692181	1.018352	1.676245	-1.349603	-1.264598	-1.758359	-1.581139
1	-0.411675	-0.260882	1.040021	0.908097	-0.380813	-0.872618	-0.637461	-0.718230	-0.651004	-0.634019	-0.733642	-0.168701	-1.126420	-0.571340	-1.171160	-0.132545	1.274785	-0.963297	-1.349603	1.404572	0.449070	0.632456
2	-0.411675	0.913196	-0.822754	-0.929699	-0.795514	0.357935	0.570540	-0.178542	1.339513	-0.147184	-0.037254	-0.688432	1.426865	-0.229679	1.290224	-0.544908	0.334530	0.280110	0.740959	-1.264598	-0.654644	-1.581139

¶

9. Dimensionality Reduction

Back to Table of Contents

In [48]:

p = PCA(n_components=3)
p.fit(data)

Out[48]:

PCA(n_components=3)

In [49]:

W = p.components_.T
W

Out[49]:

array([[ 1.13473267e-02,  1.40722208e-01, -5.07003782e-01],
       [ 2.79487089e-01,  1.78229685e-01, -7.55935183e-02],
       [-2.46222761e-01,  5.22148135e-03,  2.82798451e-01],
       [-9.87087936e-02,  4.62106956e-01, -1.51594933e-01],
       [ 3.52658544e-03,  1.62575982e-02,  3.57603333e-02],
       [ 2.55717056e-01,  2.09388871e-01, -1.16620020e-01],
       [ 2.38397768e-01,  1.10738556e-02,  2.55126443e-01],
       [ 2.85462034e-01,  9.91142961e-03,  7.66114059e-02],
       [ 2.48709872e-01,  2.64033778e-04,  2.53718430e-01],
       [ 2.37301862e-01,  2.16100946e-02,  2.56815437e-01],
       [ 1.88380393e-01,  1.23076136e-01,  1.99232637e-01],
       [-7.82601067e-02,  3.48738784e-01,  1.52531891e-01],
       [ 1.67559836e-01,  2.96777065e-01,  2.24132613e-02],
       [ 2.77349209e-01,  1.05968036e-01,  1.67395764e-02],
       [ 2.41542769e-01,  2.05476692e-01, -6.85578463e-03],
       [-2.25949470e-01,  4.61052968e-02,  9.13280992e-02],
       [ 3.84646771e-02,  2.34783479e-01, -4.28443726e-01],
       [ 3.20099392e-01,  1.33707921e-01,  3.77299448e-02],
       [-2.75762047e-02,  1.25511392e-01,  3.00669682e-01],
       [-2.48087479e-01,  3.39317455e-01,  9.25245757e-02],
       [-2.19729538e-01,  3.46882567e-01,  2.35257576e-01],
       [-2.42808159e-01,  2.92281800e-01,  8.17617105e-02]])

In [50]:

pd.DataFrame(W, index=data.columns, columns=['W1','W2','W3'])

Out[50]:

	W1	W2	W3
Education	0.011347	0.140722	-0.507004
Income	0.279487	0.178230	-0.075594
Kidhome	-0.246223	0.005221	0.282798
Teenhome	-0.098709	0.462107	-0.151595
Recency	0.003527	0.016258	0.035760
MntWines	0.255717	0.209389	-0.116620
MntFruits	0.238398	0.011074	0.255126
MntMeatProducts	0.285462	0.009911	0.076611
MntFishProducts	0.248710	0.000264	0.253718
MntSweetProducts	0.237302	0.021610	0.256815
MntGoldProds	0.188380	0.123076	0.199233
NumDealsPurchases	-0.078260	0.348739	0.152532
NumWebPurchases	0.167560	0.296777	0.022413
NumCatalogPurchases	0.277349	0.105968	0.016740
NumStorePurchases	0.241543	0.205477	-0.006856
NumWebVisitsMonth	-0.225949	0.046105	0.091328
Age	0.038465	0.234783	-0.428444
Spent	0.320099	0.133708	0.037730
Living_With	-0.027576	0.125511	0.300670
Children	-0.248087	0.339317	0.092525
Family_Size	-0.219730	0.346883	0.235258
Is_Parent	-0.242808	0.292282	0.081762

In [51]:

p.explained_variance_

Out[51]:

array([8.27465625, 2.92091448, 1.43059731])

In [52]:

p.explained_variance_ratio_

Out[52]:

array([0.3759507 , 0.13270882, 0.06499775])

In [53]:

pd.DataFrame(p.explained_variance_ratio_, index=range(1,4), columns=['Explained Variability'])

Out[53]:

	Explained Variability
1	0.375951
2	0.132709
3	0.064998

In [54]:

p.explained_variance_ratio_.cumsum()

Out[54]:

array([0.3759507 , 0.50865952, 0.57365727])

In [55]:

sns.barplot(x = list(range(1,4)), y = p.explained_variance_, palette = 'GnBu_r')
plt.xlabel('i')
plt.ylabel('Lambda i');

In [56]:

data_PCA = pd.DataFrame(p.transform(data), columns=(['col1', 'col2', 'col3']))

In [57]:

data_PCA.describe().T

Out[57]:

	count	mean	std	min	25%	50%	75%	max
col1	2212.0	0.000000e+00	2.876570	-5.915031	-2.548037	-0.784933	2.418554	7.441146
col2	2212.0	4.497106e-17	1.709068	-4.398445	-1.343427	-0.133319	1.243046	6.248175
col3	2212.0	-1.606109e-17	1.196079	-3.542882	-0.864346	-0.015402	0.825659	5.035511

In [58]:

x = data_PCA['col1']
y = data_PCA['col2']
z = data_PCA['col3']

fig = plt.figure(figsize=(13,8))
ax = fig.add_subplot(111, projection='3d')
ax.scatter(x,y,z, c='darkred', marker='o')
ax.set_title('A 3D Projection of Data In the Reduced Dimension')
plt.show()

¶

10. Clustering

Back to Table of Contents

Elbow Method to determine the number of clusters to be formed.

In [59]:

Elbow_M = KElbowVisualizer(KMeans(), k=10)
Elbow_M.fit(data_PCA)
Elbow_M.show();

In [60]:

AC = AgglomerativeClustering(n_clusters=4)
# fit model and predict clusters
yhat_AC = AC.fit_predict(data_PCA)
data_PCA['Clusters'] = yhat_AC
#Adding the Clusters feature to the orignal dataframe.
data['Clusters'] = yhat_AC
data_old['Clusters'] = yhat_AC

In [61]:

fig = plt.figure(figsize=(13,8))
ax = plt.subplot(111, projection='3d', label='bla')
ax.scatter(x, y, z, s=40, c=data_PCA['Clusters'], marker='o', cmap='Set1_r')
ax.set_title('Clusters')
plt.show()

¶

11. Clusters Analysis and Customers Profiling

Back to Table of Contents

In [62]:

pal = ['gold','#cc0000', '#ace600','#33cccc']
plt.figure(figsize=(13,8))
pl = sns.countplot(x=data['Clusters'], palette= pal)
pl.set_title('Distribution Of The Clusters')
plt.show()

In [63]:

plt.figure(figsize=(13,8))
pl = sns.scatterplot(data=data_old, x=data_old['Spent'], y=data_old['Income'], hue=data_old['Clusters'], palette= pal)
pl.set_title("Cluster's Profile Based on Income and Spending")
plt.legend();

Income vs spending plot shows the clusters pattern

group 1: high spending & average income
group 0: low spending & low income
group 3: low spending & average income
group 2: high spending & high income

In [76]:

plt.figure(figsize=(13,8))
pl = sns.swarmplot(x=data_old['Clusters'], y=data_old['Spent'], color="#CBEDDD", alpha=0.7)
pl = sns.boxenplot(x=data_old['Clusters'], y=data_old['Spent'], palette=pal)
plt.show();

From the above plot, it can be clearly seen that cluster 2 is our biggest set of customers closely followed by cluster 1. We can explore what each cluster is spending on for the targeted marketing strategies.

Plotting count of total campaign accepted.

In [75]:

data_old['Total_Promos'] = data_old['AcceptedCmp1']+ data_old['AcceptedCmp2']+ data_old['AcceptedCmp3']+ data_old['AcceptedCmp4']+ data_old['AcceptedCmp5']

plt.figure(figsize=(13,8))
pl = sns.countplot(x=data_old['Total_Promos'], hue=data_old['Clusters'], palette= pal)
pl.set_title('Count Of Promotion Accepted')
pl.set_xlabel('Number Of Total Accepted Promotions')
plt.legend(loc='upper right')
plt.show();

There has not been an overwhelming response to the campaigns so far. Very few participants overall. Moreover, no one part take in all 5 of them. Perhaps better-targeted and well-planned campaigns are required to boost sales.

Plotting the number of deals purchased

In [77]:

plt.figure(figsize=(13,8))
pl=sns.boxenplot(y=data_old['NumDealsPurchases'],x=data_old['Clusters'], palette= pal)
pl.set_title('Number of Deals Purchased');

Unlike campaigns, the deals offered did well. It has best outcome with cluster 1 and cluster 3. However, our star customers cluster 2 are not much into the deals. Nothing seems to attract cluster 0 overwhelmingly

In [80]:

Personal = ['Kidhome', 'Teenhome', 'Age', 'Children', 'Family_Size', 'Is_Parent', 'Education', 'Living_With']

for i in Personal:
    plt.figure(figsize=(13,8))
    sns.jointplot(x=data_old[i], y=data_old['Spent'], hue=data_old['Clusters'], kind='kde', palette=pal);

<Figure size 936x576 with 0 Axes>

<Figure size 936x576 with 0 Axes>

<Figure size 936x576 with 0 Axes>

<Figure size 936x576 with 0 Axes>

<Figure size 936x576 with 0 Axes>

<Figure size 936x576 with 0 Axes>

<Figure size 936x576 with 0 Axes>

<Figure size 936x576 with 0 Axes>

About Cluster 0:

The majority of these people are parents
At max have 3 members in the family
They majorly have one kid and typically not tennagers
Relatively younger

About Cluster 1:

Definitely a parent
At max have 4 members in the family and at least 2
Most have a teeanger in home
Single parents are a subset of this group
Relatively older

About Cluster 2:

Definitely not a parent
At max are only 2 members in the family.
A slight majority of couples over single people
Span all ages
high income and high spending

About Cluster 3:

Definitely a parent
At max have 5 members in the family and at least 2
Majority of them have a teenager at home
Relatively older