The number of people who do not use credit cards nowadays is very small. But how efficiently do we use it? To us, this question may sound like a social question. But for banks, this question refers to the risk they take or the opportunities they can seize. In order to evaluate these risks and opportunities, they need to analyze the situation of their customers well. That's why customer segmentation is essential.
Our aim in this project, which examines the status of customers according to credit card behaviors, is to determine which customer we should approach and how. To put it another way, in this case we will develop a customer segmentation to define marketing strategy.
Our sequence of actions is as follows: Answering questions such as how customers spend, who gets more cash, who buys more expensive, who buys more in installments etc. According to the answers to these questions we asked, we can take actions such as which customer's limit will be increased or decreased, which customers should we inform about the installment opportunities.
Using the K-Means algorithm, I made a clustering by examining the spending of customers with their credit cards.
The explanations of the codes are also written as a comment line in detail, but I will write the operations I have done.
After defining the libraries we use, we read our file. I defined the reading patterns for the colab environment and for the local environment.
To familiarize myself with the dataset, I took a look at the variable names and types.
First of all, I examined values such as mean value, standard deviation, min max with the .describe() function and saw that there may be outliers here.
After I filled in the missing values with their mean, I started to detect the outliers.
As a next step, I plotted graphs between variables of my choice to observe outliers.
I printed out the visibly obvious outliers and their percentages separately for each column with a for loop.
Deleting it would cause a huge data loss as each column had a large amount of outliers. So I made ranges to deal with outliers
After standardizing the data with StandardScaler, I moved on to the modeling phase.
I observed the optimal clustering number with the Elbow method.
I ran the K-Means model with the appropriate hyperparameters and fitted our model.
I decided 6 clusters. Also I decided n_init for initialize cluster number, max_iter for how many iterations we will do, random_state is random state as all you know :)
I interpreted with FacetGrid by observing the relations of clusters with variables.
The classes I clustered are as follows
-
Cluster5 People with average to high credit limit who make all type of purchases
-
Cluster4 This group has more people with due payments who take advance cash more often
-
Cluster3 Less money spenders with average to high credit limits who purchases mostly in installments
-
Cluster2 People with high credit limit who take more cash in advance
-
Cluster1 High spenders with high credit limit who make expensive purchases
-
Cluster0 People who don't spend much money and who have average to high credit limit
Before using PCA to transform data to 2 dimensions for visualization, first we took cosine distance before PCA because cosine distance kills the magnitude of vectors and focuses on relationships only.
Finally after PCA, I visualized our data reduced to x and y tags, expressing the clusters with different colors, grouping them according to their labels with the .groupby function to observe.
-
pandas(load and manipulate data and for One-Hot Encoding) -
numpy(calculate the mean and standard deviation) -
seaborn(helping with some visualization techniques) -
matplotlib.pyplot(some graphs) -
matplotlib.ticker(for specifying the axes thick format) -
sklearn.preprocessing, StandartScaler (scaling) -
sklearn.cluster, KMeans (k-means algorithm library) -
sklearn.decomposition, PCA (applying principal component analysis) -
sklearn.metrics.pairwise, cosine_similarity (normalised dot product between two vectors (preparing for PCA)) -
io(reading files for all systems)
The sample Dataset summarizes the usage behavior of about 9000 active credit card holders during the last 6 months. The file is at a customer level with 18 behavioral variables.
Following is the Data Dictionary for Credit Card dataset :
CUSTID : Identification of Credit Card holder (Categorical)
BALANCE : Balance amount left in their account to make purchases (
BALANCEFREQUENCY : How frequently the Balance is updated, score between 0 and 1 (1 = frequently updated, 0 = not frequently updated)
PURCHASES : Amount of purchases made from account
ONEOFFPURCHASES : Maximum purchase amount done in one-go
INSTALLMENTSPURCHASES : Amount of purchase done in installment
CASHADVANCE : Cash in advance given by the user
PURCHASESFREQUENCY : How frequently the Purchases are being made, score between 0 and 1 (1 = frequently purchased, 0 = not frequently purchased)
ONEOFFPURCHASESFREQUENCY : How frequently Purchases are happening in one-go (1 = frequently purchased, 0 = not frequently purchased)
PURCHASESINSTALLMENTSFREQUENCY : How frequently purchases in installments are being done (1 = frequently done, 0 = not frequently done)
CASHADVANCEFREQUENCY : How frequently the cash in advance being paid
CASHADVANCETRX : Number of Transactions made with "Cash in Advanced"
PURCHASESTRX : Numbe of purchase transactions made
CREDITLIMIT : Limit of Credit Card for user
PAYMENTS : Amount of Payment done by user
MINIMUM_PAYMENTS : Minimum amount of payments made by user
PRCFULLPAYMENT : Percent of full payment paid by user
TENURE : Tenure of credit card service for user
Dataset link: https://www.kaggle.com/arjunbhasin2013/ccdata
Different actions can be taken according to the 6 clusters we have separated. For example,(for Cluster3) we can encourage users who shop in installments by increasing their special installment options or (for Cluster2), customers who withdraw and spend cash can be provided with low commission cash via credit card or (for Cluster5) we can increase their loyalty with special offers for people with high limits and all kinds of spending.