Credit Card Clients

Radhey Shyam
Posted on May 1, 2016

Contributed by Radhey Shyam. He  is currently in the NYC Data Science Academy 12 week full time Data Science Bootcamp program taking place between April 11th to July 1st, 2016. This post is based on his first class project - R visualization (due on the 2nd week of the program).

According to Business Insider, credit card industry in U.S. was $4 trillion and in comparison, size of the total U.S. economy was $17.4 trillion in 2014. The credit card transactions has been increasing for years and there were $26.2 billion transactions in 2012.

Dataset

The dataset for this project consists of default payments from a major credit card company in Taiwan and is available at UCI Machine Learning Repository. The data set consists of 30,000 instances and 24 attributes consisting of gender, education profile, marital status,age, history of statement balance, payment status and binary status of default ( 1 or 0).

First of all, I read data from xls file and created data frame to store the data and created df table to store the data.For this, I wrote the following code:

https://gist.github.com/rshyam1/b4c888a94e1d67ebc5bac4e63db543ed

Next, I created gender contingency table for default (1) and non-default(0) categories:

https://gist.github.com/rshyam1/1ad5c82bb3ccb8576c5d486b8a154291

gender_table

As we can see from the table, the data set consists of approximately 78 % (23364) individuals with default status and 22 % (6634) having non default status.

Then, I wanted to visualize the gender status among two categories of default and non-default.The default status is plotted on x-axis and number of individuals are plotted on y-axis and gender status was dodged on the x-axis.The numerical gender values in the dataset were changed to string values of "male" and "female".

https://gist.github.com/rshyam1/f1107bac4aaa493bd61b689cb6c273e3

gender_profile_800

As, we can see from the above bar chart, number of females is higher than males for both default and no default payment categories, but we cannot see their respective ratios. Next, I plotted another bar chart to show the gender ratio and the fill color was based on gender status.

gender_ratios

Next, I thought of visualizing education profile of customers . I changed the education numeric value in the dataset to strings and created bar plots for education as factors for fill and default status on x-axis.

https://gist.github.com/rshyam1/f46ca1ef641f65deacbd422ae39bf10b

education_bar

We can see from above bar chart that number of customers having university education is highest,   followed by customers having graduate school and high school in both the categories. To compare their relative ratios, I made another bar graph and this time, education factor was used as fill color.

education_bar_fill

As we can see from above bar graph, ratio of customers having graduate school is higher in default payment  category as compared to no-default and ratio of customers having university education is higher in No Default Payment than Default category.

Next, I wanted to compare age distribution of customers in Default Status vs no Default Payment.
For that purpose, I thought that violin plot will be best suited to show the age density among two categories.

https://gist.github.com/rshyam1/4cada454e89641c7cd0424a3d3fd9f41

age_violin_plot

From above age violin plot, we can see that the age peaks around 28 years in the default payment category and then it age goes on decreasing after the peak. To clearly see the peaks, I plotted the age density plots below.

age_density

Now, we can see clearly see that age peaks around 28-29 years in default category and it has lower peak around 27-28 years in no default category.

To compare the marital status among two categories, I first converted the numerical values in the data to their respective strings and then created the bar chart

https://gist.github.com/rshyam1/e37f2f7b7dde30a64bf387bdfea0cf9f

marital_status
The above bar graph shows the ratio of individuals having single status is higher in default status vs non-default.

Next, I thought of calculating monthly default rate for the dataset to see its trend during observation period (April 2005 to September 2005)  . First, I took the accepted definition of default as  the individuals who have not paid for mare than two months. For calculating the monthly default rate,I divided that number by the total.

https://gist.github.com/rshyam1/1c841264a26708d32ebd66b7c7df594a

default_rate_scatter

The above scatter graph shows the default rate is linearly increasing from April 2005 to August 2005 and then slowing down in September 2005.

Finally, to find the correlation between Credit Limit, Bill amounts and payment status, I created the following correlation plot.

core_plot

The above correlation plot shows a strong correlation between credit limit and billing amounts from BILL_AMT6(April 2005)  to  BILL_AMT1 ( Sep. 2005) and no correlation between credit limit and payment status during that period.

In conclusion, data exploration of credit card default dataset shows 1) larger percentage of females than males in default payment category 2) percentage of  customers having graduate school degree is higher in default payment  category 3) individuals having single status have higher percentage of default than married 4) age peaks around 28-29 years in default payment category and has lower peak around around 27-28 years in no default payment category as verified from violin and density profile.3) default rate increases during the data collection from April 2005 to September 2005.
In the future, I would like to test different machine learning algorithms to predict the default status probabilities and benchmark  their performances.

About Author

Radhey Shyam

Radhey Shyam

Radhey Shyam received his Ph.D. in Physics from Clemson University, USA and M.S. (Computer Science & Engineering) from GJU of Science & Technology, India. Radhey Shyam did his master's thesis on Network Monitor System, where he applied innovative...
View all posts by Radhey Shyam >

Related Articles

Leave a Comment

Avatar
Fredrik Davéus August 14, 2016
Hi, I was just looking at this data set and found this blog post. If I interpret the original data description and also your blog text correctly a value of "1" in the column "default payment next month" means that the person is in default, and a value of "0" in the same column means that the person is not in default. Then I conclude that there is 6636 individuals in default and the rest in non-default. However, your first table seems to suggest that the opposite is the case. Please clarify any mistake on my part. Thanks/Regards, Fredrik
Avatar
Emilio June 21, 2016
You made some decent points there. I looked on the web for the issue and found most individuals will go along with with your site.

View Posts by Categories


Our Recent Popular Posts


View Posts by Tags

#python #trainwithnycdsa 2019 airbnb Alex Baransky alumni Alumni Interview Alumni Reviews Alumni Spotlight alumni story Alumnus API Application artist aws beautiful soup Best Bootcamp Best Data Science 2019 Best Data Science Bootcamp Best Data Science Bootcamp 2020 Best Ranked Big Data Book Launch Book-Signing bootcamp Bootcamp Alumni Bootcamp Prep Bundles California Cancer Research capstone Career Career Day citibike clustering Coding Course Demo Course Report D3.js data Data Analyst data science Data Science Academy Data Science Bootcamp Data science jobs Data Science Reviews Data Scientist Data Scientist Jobs data visualization Deep Learning Demo Day Discount dplyr employer networking feature engineering Finance Financial Data Science Flask gbm Get Hired ggplot2 googleVis Hadoop higgs boson Hiring hiring partner events Hiring Partners Industry Experts Instructor Blog Instructor Interview Job Job Placement Jobs Jon Krohn JP Morgan Chase Kaggle Kickstarter lasso regression Lead Data Scienctist Lead Data Scientist leaflet linear regression Logistic Regression machine learning Maps matplotlib Medical Research Meet the team meetup Networking neural network Neural networks New Courses nlp NYC NYC Data Science nyc data science academy NYC Open Data NYCDSA NYCDSA Alumni Online Online Bootcamp Open Data painter pandas Part-time Portfolio Development prediction Prework Programming PwC python python machine learning python scrapy python web scraping python webscraping Python Workshop R R language R Programming R Shiny r studio R Visualization R Workshop R-bloggers random forest Ranking recommendation recommendation system regression Remote remote data science bootcamp Scrapy scrapy visualization seaborn Selenium sentiment analysis Shiny Shiny Dashboard Spark Special Special Summer Sports statistics streaming Student Interview Student Showcase SVM Switchup Tableau team TensorFlow Testimonial tf-idf Top Data Science Bootcamp twitter visualization web scraping Weekend Course What to expect word cloud word2vec XGBoost yelp