Customer Segmentation - Using k-means

About: Customer Segmentation is a popular application of unsupervised learning. Using clustering, identify segments of customers to target the potential user base. They divide customers into groups according to common characteristics like gender, age, interests, and spending habits so they can market to each group effectively.

Use K-means clustering and also visualize the gender and age distributions. Then analyze their annual incomes and spending scores.

#importing required libraries
import numpy as np
import pandas as pd
import seaborn as sns
import matplotlib.pyplot as plt
from sklearn.cluster import KMeans

import warnings
warnings.simplefilter(action='ignore', category=FutureWarning)

#reading data using pands to a dataframe and printing its head values
customer = pd.read_csv('Customer-Segmentation.csv')
customer.head()

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

There are 5 columns CustomerID, Gender, Age, Annual Income and Spending Score in our dataframe 'customer'

#checking size of data 
customer.shape

(200, 5)

We have a data set with 200 rows and 5 columns.

#checking dataframe for any NULL values
customer.isnull().sum()

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

It clearlly shows that there is no NULL value present in our dataframe.

#Getting 5 point summary of our dataframe
customer.describe()

	CustomerID	Age	Annual Income (k$)	Spending Score (1-100)
count	200.000000	200.000000	200.000000	200.000000
mean	100.500000	38.850000	60.560000	50.200000
std	57.879185	13.969007	26.264721	25.823522
min	1.000000	18.000000	15.000000	1.000000
25%	50.750000	28.750000	41.500000	34.750000
50%	100.500000	36.000000	61.500000	50.000000
75%	150.250000	49.000000	78.000000	73.000000
max	200.000000	70.000000	137.000000	99.000000

We got values like mean, std deviation, min, max, Q1, Q2 and Q3 for all attributes.

#applying grid to all our plots for better visuals
sns.set(style="whitegrid")

Visualizing various Distributions

# distribution plot for 'Age'
sns.distplot(customer['Age'],color= 'green',bins=20)
plt.title('Age distribution plot', fontsize = 15)
plt.xlabel('Age', fontsize = 12)
plt.ylabel('Count', fontsize = 12)
plt.show()

This shows that our data has customer ranges from 10 years to 80 years.

# count plot for 'Age'
plt.figure(figsize=(20,8))
sns.countplot(customer['Age'])
plt.title('Age count plot', fontsize = 15)
plt.xlabel('Age', fontsize = 12)
plt.ylabel('Count', fontsize = 12)
plt.show()

This plot is more clear view on counting customer based on their Age. Also we can see that 11 customers are 32 years old which is the most value count.

# box plot for 'Age'
sns.boxplot(customer['Age'])
plt.title('Age box plot', fontsize = 15)
plt.xlabel('Age', fontsize = 12)
plt.show()

Based on 5 point summary we can get a clear picture of various aspect of customer based on their age.

# violin plot for 'Age'
sns.violinplot(y = 'Age' , x = 'Gender' , data = customer)
plt.title('Age violin plot', fontsize = 15)
plt.xlabel('Gender', fontsize = 12)
plt.ylabel('Age', fontsize = 12)
plt.show()

This violin plot shows that we have higher number of female customer who belongs to age group of 30 years.

# distribution plot for 'Annual Income'
sns.distplot(customer['Annual Income (k$)'],color= 'blue',bins=20)
plt.title('Annual Income distribution plot', fontsize = 15)
plt.xlabel('Annual Income', fontsize = 12)
plt.ylabel('Count', fontsize = 12)
plt.show()

This shows that our data has customer ranges from income of 0k to 150k.

# count plot for 'Annual Income'
plt.figure(figsize=(20,8))
sns.countplot(customer['Annual Income (k$)'])
plt.title('Annual Income', fontsize = 15)
plt.xlabel('Annual Income', fontsize = 12)
plt.ylabel('Count', fontsize = 12)
plt.show()

This plot is more clear view on counting customer based on their Income. Also we can see that 12-12 customers are 54 years and 78 years old which is the most value count.

# box plot for 'Annual Income'
sns.boxplot(customer['Annual Income (k$)'])
plt.title('Annual Income box plot', fontsize = 15)
plt.xlabel('Annual Income', fontsize = 12)
plt.show()

Based on 5 point summary we can get a clear picture of various aspect of customer based on their income.

# violin plot for 'Annual Income'
sns.violinplot(y = 'Annual Income (k$)' , x = 'Gender' , data = customer)
plt.title('Annual Income violin plot', fontsize = 15)
plt.xlabel('Gender', fontsize = 12)
plt.ylabel('Annual Income', fontsize = 12)
plt.show()

This violin plot shows that we have higher number of male customer who have more income.

# distribution plot for 'Spending Score'
sns.distplot(customer['Spending Score (1-100)'],color= 'orange',bins=20)
plt.title('Spending Score distribution plot', fontsize = 15)
plt.xlabel('Spending Score', fontsize = 12)
plt.ylabel('Count', fontsize = 12)
plt.show()

This shows that our data has customer ranges from with -20 to 120 spending score.

# count plot for 'Spending Score'
plt.figure(figsize=(20,8))
sns.countplot(customer['Spending Score (1-100)'])
plt.title('Spending Score count plot', fontsize = 15)
plt.xlabel('Spending Score', fontsize = 12)
plt.ylabel('Count', fontsize = 12)
plt.show()

This plot is more clear view on counting customer based on their Age. Also we can see that 8 customers are 42 years old which is most value count.

# box plot for 'Spending Score'
sns.boxplot(customer['Spending Score (1-100)'])
plt.title('Spending Score', fontsize = 15)
plt.xlabel('Spending Score', fontsize = 12)
plt.show()

Based on 5 point summary we can get a clear picture of various aspect of customer based on their spending score.

# violin plot for 'Spending Score'
sns.violinplot(y = 'Spending Score (1-100)' , x = 'Gender' , data = customer)
plt.title('Spending Score', fontsize = 15)
plt.xlabel('Gender', fontsize = 12)
plt.ylabel('Spending Score', fontsize = 12)
plt.show()

This violin plot shows that we have higher number of female customer who have mostly spending score around 50 .

# count plot for 'Gender'
sns.countplot(x='Gender', data=customer)
plt.title('Spending Score count plot', fontsize = 15)
plt.xlabel('Gender', fontsize = 12)
plt.ylabel('Count', fontsize = 12)
plt.show()

This plot clearly shows that we have more female customer compare to male customers.

# heatmap to show correlation of various Attributes
sns.heatmap(customer.corr(), annot = True)

<AxesSubplot:>

From this plot we got that income and spending score correlates to each other with a good score. But age and spending score does not correlates efficiently.

Cluster based on Annual Income and Spending Score

plt.figure(figsize=(15,8))
sns.scatterplot(customer['Annual Income (k$)'], customer['Spending Score (1-100)'],hue=customer['Gender'],
                palette= ['red','green'] ,alpha=0.6)
plt.title('Distribution of Gender based on Annual Income and Spending Score', fontsize = 15)
plt.xlabel('Annual Income', fontsize = 12)
plt.ylabel('Spending Score', fontsize = 12)
plt.show()

This scatter plot show the distribution of customers based on their income, spending score and gender. And we can see customer cluster clearlly in this plot.

Income_Spend = customer[['Annual Income (k$)' , 'Spending Score (1-100)']].iloc[: , :].values
from sklearn.cluster import KMeans

wcss = []
for i in range(1, 11):
    km = KMeans(n_clusters = i, init = 'k-means++', max_iter = 300, n_init = 10, random_state = 0)
    km.fit(Income_Spend)
    wcss.append(km.inertia_)

plt.figure(figsize=(15,8))
plt.plot(range(1, 11), wcss)
plt.title('The Elbow Method', fontsize = 15)
plt.xlabel('No. of Clusters', fontsize = 12)
plt.ylabel('wcss', fontsize = 12)
plt.show()   

This elbow method show a low slope line after 5 number of cluster so we can take 5 as optimum number of cluster.

km = KMeans(n_clusters = 5, init = 'k-means++', max_iter = 300, n_init = 10, random_state = 0)
y_means = km.fit_predict(Income_Spend)

plt.figure(figsize=(15,8))
plt.scatter(Income_Spend[y_means == 0, 0], Income_Spend[y_means == 0, 1], s = 100, c = 'pink', label = 'Average')
plt.scatter(Income_Spend[y_means == 1, 0], Income_Spend[y_means == 1, 1], s = 100, c = 'yellow', label = 'Spenders')
plt.scatter(Income_Spend[y_means == 2, 0], Income_Spend[y_means == 2, 1], s = 100, c = 'cyan', label = 'Best')
plt.scatter(Income_Spend[y_means == 3, 0], Income_Spend[y_means == 3, 1], s = 100, c = 'magenta', label = 'Low Budget')
plt.scatter(Income_Spend[y_means == 4, 0], Income_Spend[y_means == 4, 1], s = 100, c = 'orange', label = 'Saver')
plt.scatter(km.cluster_centers_[:,0], km.cluster_centers_[:, 1], s = 50, c = 'blue' , label = 'centeroid')
plt.legend()
plt.title('Customere Segmentation using Annual Income and Spending Score', fontsize = 15)
plt.xlabel('Annual Income', fontsize = 12)
plt.ylabel('Spending Score', fontsize = 12)
plt.show()

Based on the above clustering we can clearly say that there are five cluster segments present based on customers' Annual Income and Spending Score. We named them as Low budget, Spenders, Average, Savers, and Best.

Cluster based on Age and Spending Score

plt.figure(figsize=(15,8))
sns.scatterplot(customer['Age'], customer['Spending Score (1-100)'],hue=customer['Gender'],palette= ['red','green'] ,alpha=0.6)
plt.title('Distribution of Gender based on Age and Spending Score', fontsize = 15)
plt.xlabel('Age', fontsize = 12)
plt.ylabel('Spending Score', fontsize = 12)
plt.show()

This scatter plot show the distribution of customers based on their age, spending score and gender. And we can clearly observe that aged people don't have higher spending score.

Age_Spend = customer[['Age' , 'Spending Score (1-100)']].iloc[: , :].values
wcss = []
for i in range(1, 11):
    km = KMeans(n_clusters = i, init = 'k-means++', max_iter = 300, n_init = 10, random_state = 0)
    km.fit(Age_Spend)
    wcss.append(km.inertia_)

plt.figure(figsize=(15,8))
plt.plot(range(1, 11), wcss)
plt.title('The Elbow Method', fontsize = 15)
plt.xlabel('No. of Clusters', fontsize = 12)
plt.ylabel('wcss', fontsize = 12)
plt.show()  

This elbow method show a low slope line after 4 noumber of cluster so we can take 4 as optimum number of cluster.

km = KMeans(n_clusters = 4, init = 'k-means++', max_iter = 300, n_init = 10, random_state = 0)
ymeans = km.fit_predict(Age_Spend)

plt.figure(figsize=(15,8))
plt.scatter(Age_Spend[ymeans == 0, 0], Age_Spend[ymeans == 0, 1], s = 100, c = 'pink', label = 'Young Targets' )
plt.scatter(Age_Spend[ymeans == 1, 0], Age_Spend[ymeans == 1, 1], s = 100, c = 'orange', label = 'Regular Customers')
plt.scatter(Age_Spend[ymeans == 2, 0], Age_Spend[ymeans == 2, 1], s = 100, c = 'lightgreen', label = 'Usual Customer')
plt.scatter(Age_Spend[ymeans == 3, 0], Age_Spend[ymeans == 3, 1], s = 100, c = 'red', label = 'Old Targets')
plt.scatter(km.cluster_centers_[:, 0], km.cluster_centers_[:, 1], s = 50, c = 'black')
plt.legend()
plt.title('Customere Segmentation using Annual Income and Spending Score', fontsize = 15)
plt.xlabel('Age', fontsize = 12)
plt.ylabel('Spending Score', fontsize = 12)
plt.show()

Based on the above clustering we can clearly say that there are four cluster segments present based on customers' Age and Spending Score. We named them as Regular Customers, Usual Customer, Young Targets, and Old Targets.

And, Project is over!!!

Completed by: Ganpat Patel

Email: ganpat.patel.012@gmail.com