Credit Card Fraud Analysis 

Yash Saini, Chinmay Agrawal and Udayakumar

Agenda
---------------------------------

 Objective
 Background
 Key Insights
 Cost Benefit Analysis
 Recommendations
 Appendix
 Data attributes
 Data Imbalance Handling
 Models
 References
Objective
Need to recommend a credit card fraud detection system for Finex, a
leading financial provider
in US, to save incurred cost due to fraudulent transactions with better
customer experience.
Background
Finex is a leading financial service provider based out of Florida, US. It offers a wide range of products and
business services to customers through different channels, ranging from in-person banking and ATMs to online banking.
Over the last few years, Finex has observed that a significantly large number of unauthorised transactions are being
made, due to which the bank has been facing a huge revenue and profitability crisis. Many customers have been
complaining about unauthorised transactions being made through their credit/debit card It has been reported that
fraudsters use stolen/lost cards and hack private systems to access the personal and sensitive data of many cardholders.
They also indulge in ATM skimming at various POS terminals such as gas stations, shopping malls, and ATMs that do
not send alerts or do not have OTP systems through banks. Such fraudulent activities have been reported to happen
during non-peak and odd hours of the day leaving no room for suspicion.

In most cases, customers get to know of such unauthorised transactions happening through their cards quite leas they
are unaware of such ongoing credit card frauds or they do not monitor their bank account activities closely. This has led
to late complaint registration with Finex and by the time the case is flagged fraudulent, the bank incurs heavy losses and
ends up paying the lost amount to the cardholders.

Now, Finex is also not really equipped with the latest financial technologies, and it is becoming difficult for thebank to
track these data breaches on time to prevent further losses. The Branch Manager is worried about the ongoing situation
and wants to identify the possible root causes and action areas to come up with a solution that would help the bank
generate high revenue with minimal losses.
Appendix
 Data Imbalance Handling
ADASYN: Adaptive Synthetic (ADASYN) is based on the idea of adaptively generating minority data
samples according to their distributions using K nearest neighbour. The algorithm adaptively updates the
distribution and there are no assumptions made for the underlying distribution of the data. The algorithm
uses Euclidean distance for KNN Algorithm. The key difference between ADASYN and SMOTE is
that the former uses a density distribution, as a criterion to automatically decide the number of synthetic
samples that must be generated for each minority sample by adaptively changing the weights of the different
minority samples to compensate for the skewed distributions. The latter generates the same number of
synthetic samples for each
Results on Unseen Data by Model Predictors :
XGBoost - ADASYN Modelled Predictors: XGBoost - SMOTE Modelled Predictors:

Decision Tree - SMOTE Modelled Predictors: Decision Tree - ADASYN Modelled Predictors:

 Random Forest - SMOTE Modelled Predictors: Random Forest - ADASYN Modelled Predictors:
Recommendations

 We recommend adoption of Decision trees ! with ADASYN data

generation technique as the data is highly imbalanced.

 It is recommended to use model that can detect maximum fraudulent

transaction whereas False Positive cases which are like not fraudulent
but model predicted as fraudulent should be ok as the cost is less. But
this may increase customer dissatisfaction as well, for having frequent
validations.
References :
1. Jupyter Notebook : Credit_Card_Fraud_Capstone_Project.ipynb
2. PDF File : Rising_Frauds.pdf
3. PDF File : Credit Card Fraud Analysis.pdf
4. Excel File : Cost+Benefit+Analysis.xlsx
THANK YOU