Determine Customer Spending Patterns Using K-means Clustering

Table of Contents

• Determine Customer Spending Patterns Using K-means Clustering
  • Table of Contents
  • Project Objective
  • Importing Python Packages
  • Importing the Dataset
  • Using the Elbow Method to Find the Optimal Number of Clusters
  • Training the Final Model and Visualizing Clusters

Project Objective

Analyze customer spending habits using K-means clustering on annual income and spending score data. The goal is to identify distinct customer groups for targeted marketing and personalized services. By visualizing the clusters, I aim to gain insights into spending patterns and relationships between income and behavior.

KMeans Clustering Steps:

1. Choose the number of K clusters: Determine the optimal number of clusters (K) using the Elbow method or another heuristic.
2. Select random K points as centroids: Initialize K centroids randomly.
3. Assign each data point to the closest centroid: Form K clusters by assigning each data point to its nearest centroid.
4. Compute new centroids: Calculate the mean of the data points in each cluster to find the new centroid.
5. Reassign data points: Reassign each data point to the new closest centroid. Repeat steps 4 and 5 until the centroids no longer change significantly.

Importing Python Packages

Import necessary libraries such as pandas, numpy, matplotlib, and sklearn.

# Import python packages:
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
from sklearn.cluster import KMeans
import os

print(os.environ['OMP_NUM_THREADS'])

Importing the Dataset:

df = pd.read_csv(path_to_dataset)

Using the Elbow Method to Find the Optimal Number of Clusters

• Create a 2-dimensional array of the independent variables.
• Compute WCSS for different numbers of clusters.
• Plot the results to identify the optimal number of clusters.

# Create 2 dimensional array of our independent variables:
X = df[['Annual Income (k$)', 'Spending Score (1-100)']].values

# Create empty list to store wcss values:
wcss = []

# Loop through X array and compute optimal number of clusters using the WCSS algorithm:
for i in range(1, 11):
    # Instantiate kmeans:
    kmeans = KMeans(
        # set number of clusters:
        n_clusters = i,
        # set init:
        init = "k-means++",
        # set random state:
        random_state = 42,
    )
    # Train model:
    kmeans.fit(X)
    # Append the wcss per cluster to wcss list:
    wcss.append(kmeans.inertia_)

# Plot the results:
plt.plot(range(1, 11), wcss)
plt.title("Elbow Method")
plt.xlabel('Number of Clusters')
plt.ylabel("WCSS")
plt.tight_layout()
plt.show()

To determine the optimal number of clusters, we want to locate where the elbow occurs (biggest drop in wcss). Based on the graph, the optimal number of clusters is around 5.

Training the Final Model and Visualizing Clusters

• Instantiate the K-means model with the optimal number of clusters.
• Fit the model and predict the clusters.
• Print the cluster centers and labels.
• Visualize the clusters and centroids.

# Instantiate kmeans:
kmeans = KMeans(
    # set number of clusters using the optimal number of clusters found above:
    n_clusters = 5,
    # set init:
    init = "k-means++",
    # set random state:
    random_state = 42,
)

# Fit model:
y_pred = kmeans.fit_predict(X)

# Print the cluster centers
print("Cluster centers:")
print(kmeans.cluster_centers_)

# Print the labels of the clusters
print("Cluster labels for each data point:")
print(kmeans.labels_)

# Plot the clusters
plt.figure(figsize=(10, 6))

# Colors for different clusters
colors = ['red', 'blue', 'green', 'cyan', 'magenta']

for i in range(5):
    plt.scatter(X[y_pred == i, 0], X[y_pred == i, 1], s=100, c=colors[i], label=f'Cluster {i}')

# Plot the centroids
plt.scatter(kmeans.cluster_centers_[:, 0], kmeans.cluster_centers_[:, 1], s=300, c='yellow', label='Centroids', edgecolor='black')

# Add titles and labels
plt.title('Clusters of customers')
plt.xlabel('Annual Income (k$)')
plt.ylabel('Spending Score (1-100)')
plt.legend()
plt.tight_layout()
plt.show()

Cluster Interpretation:

• High Income, Low Spending: Customers with higher annual incomes (blue cluster) tend to have lower spending scores. This suggests that despite having more disposable income, these customers are more conservative with their spending.

• Low Income, High Spending: Customers with lower annual incomes (green and pink clusters) tend to have higher spending scores. These customers are spending a larger portion of their income, indicating a more aggressive spending behavior.

• Moderate Income and Spending: There are clusters of customers (red and cyan clusters) that fall into a more moderate range of both income and spending. These groups might represent more balanced or average spenders.

Marketing Strategy

• Targeting High-Income Customers: Develop campaigns that emphasize value and exclusivity to encourage higher spending. Since these customers have the capacity to spend more, targeted promotions, loyalty programs, or exclusive offers might incentivize them to increase their spending.

• Retention and Rewards for High Spenders with Low Income: Ensure that customers with lower incomes but higher spending scores feel valued. Implementing loyalty programs or rewards that provide discounts or benefits could help retain these valuable customers.

• Balanced Approach for Moderate Groups: Continue to offer balanced promotions that cater to the needs and preferences of the moderate income and spending groups.