Blog

Perfect Credit Card Clustering with Machine Learning Models


The top challenge faced by marketers is understanding who they are selling to. Once you know your buyer personas, you can tailor your targeting and offerings to increase their satisfaction and your revenue as a result. When you already have a pool of customers and plenty of data, it can be incredibly useful to segment them.

 

Download our Machine Learning 101 guide

 

In this blog, you'll see how we can use credit card clustering to segment customers. The data for this analysis was taken from Kaggle and we will use Mind Foundry machine learning to automatically identify the clusters.

The data

The credit card data has 17 attributes for each customer, which include the balance (credit owed by the customer), cash advance (when a customer withdraws cash using the credit card), the customer’s credit limit, minimum payment, percentage of full payments and tenure.

The data is fairly clean but has some missing values which are automatically picked up  by Mind Foundry.

 

Data Preparation

 

After applying the data preparation advice suggested by Mind Foundry, we are able to look at the histogram view of the credit card customers.

 

Data Visualization

 

Clustering

Now that we have prepared the data we are going to build a clustering model using Mind Foundry which will automatically detect the optimal number of clusters which turns out to be 8.

 

Automated Clustering

 

The Silhouette coefficient, between -1 and 1, gives an indication of how close each point in one cluster is to points in the neighbouring clusters. Values close to 1 are furthest from other clusters whereas negative points overlap with others. In an ideal situation we would expect all the points of a cluster to have Silhouette coefficients close to 1. In our case, most of the clusters seem fairly well defined.

 

Download the guide

 

Customer segmentation

Cluster 1

This first group is characterised by customers who have the lowest balance and cash advance and have a fairly high percentage of full payments. These customers are known as transactors as they pay little interest charges.

 

Interpretation of Cluster 1

 

Cluster 3

This segment is characterised by customers who have high balances and cash advances and one of the lowest purchase frequencies and percentages of full payments which indicates that they are one of the most lucrative segments for the credit card provider. These are typically known as revolvers who might be using their credit card’s as a loan.

 

Interpretation of Cluster 3

 

Cluster 7

This segment is characterised by customers with the highest credit limit and the highest percentage of full payments. These are prime customers that the credit card provider can entice to increase their spending habits by raising their credit limits even more.

 

Interpretation of Cluster 7

 

Cluster 4

This segment includes fairly new customers (low tenures) who have a low balance and cash advance. The credit card provider might encourage them to increase their activity by offering cash backs, promotions, free Uber rides etc.

 

Interpretation of Cluster 4

 Download Machine Learning 101 Guide and Become an ML Expert

Try Mind Foundry

Related articles

The Layman's Guide to the Data Science Journey

For the past five years, data science has been praised as a technology that can unlock new applications and hidden insights for organisations. However, today it is struggling to live up to ...
Read the full article
shutterstock_598348394

The Value of Machine Learning in Clinical Trials

Clinical research data plays an essential role in the pharmaceutical industry, but can also eat up resources in terms of time and money. However, by utilising machine learning in clinical trials, ...
Read the full article
Two doctors looking at clipboard while their colleagues working in medical office