Credit Card Fraud Detection: Submitted by

CREDIT
CARD
FRAUD
DETECTION
Submitted By:
Suzanna Michelle Young
Deepak R
Varsha B S
BACKGRAOUND OF
PROBLEM
• Huge revenue loss due to unauthorised transactions
during non-peak and odd hours of the day.
• Source: Stolen/lost cards, ATM skimming at various
POS terminals.
• Stopping the fraudsters/fraudulent activity is highly
impossible as they actively find alternate ways.
• Extra level authentication is complex & tedious job.
• Goal is to detect and prevent the fraudulent
transactions before the financial crisis.
ROUTE CAUSE •
•
5 WHYs
Sensitive credit card details of customers is being
ANALYSIS •
accessed.
Private systems are hacked, stolen/lost cards are
misused, ATM skimming at POS terminals.
UNDERSTANDING THE
PROBLEM & IMPACT
1. Who is involved in the process?
• Fraudsters, skimmers who get unauthorized access to a customer’s credit
card info.
2. What do they do with it?

• Commit fraudulent transactions: alter genuine cards, create counterfeit cards, steal
credit cards, fraudulent telemarketing.
3. Where do the transactions happen?

• Most of them happen online and sometimes offline.
• Online frauds are hard to detect due to difficulty in amassing evidence, time &
resource constraints.
4. When does it happen?

• Most happen at non-peak and odd hours of the day when user in-active or asleep.
5. How is business affected by this?

• As bank need to repay customer the total transaction amount in case of fraudulent
transaction
• Banks suffer from substantial financial loss, lose trust & customer’s credibility.
PROCESS
FLOW
Data Import Selecting

Build models
& Data Feature with Train, Test
with various
EDA on Train Cleaning significant Split
techniques
Data impact
Validate model Model Fitting Compare model Validate model

Cost Benefit
with Recall & on Test/unseen performance & with Recall &
analysis
AUC Data chose the best AUC
CHARACTERISTICS
OF FRAUDULENT
TRANSACTIONS
• Gender v/s fraudulent
transactions
• Day of the week v/s fraudulent
transactions
• Month v/s fraudulent
transactions
CHARACTERISTICS • Hours of day v/s transactions
OF FRAUDULENT • Age v/s fraudulent transactions
TRANSACTIONS
MODEL
PERFORMANCE ON
TRAINING DATA
Evaluation metrics: Recall, AUC
1. Why Recall?
• Most likely to predict actual Positive cases(Fraud
transactions).
• Very helpful to reduce financial loss.
• Recall = TP / (TP + FN)
• High Recall refers to low false negative score.
2. Why AUC?
• Best to measure overall performance of a binary
classification problem.
• Best to select optimal threshold value of a model.
• Plots TPR v/s FPR.
MODEL • Trade off b/w Random Forest SMOTE & Random Forest ADASYN.
PERFORMANCE
• Still choose Random Forest ADASYN why?
• Impact from 1% increase in Recall score.
ON TEST/UNSEEN • Assumption: Financial impact of 1% fraud in 1% of overall transaction

• 1% of overall transaction = 18,524 (Train+Test)
DATA • 1% of Fraudulent transactions = 0.01% overall transaction = 185 transactions.

• Avg amt for 185 transaction = (avg amount/transaction) * $185 = $98,050
COST BENEFIT ANALYSIS
Cost Benefit Analysis

S. No Questions Answer
a Average number of transactions per month 77183.00
b Average number of fraudulent transaction per month 402.00
c Average amount per fraud transaction 530.00
Cost Benefit Analysis

S. No Questions Answer
1 Cost incurred per month before the model was deployed = (avg no of fraud transaction) * (avg amt / fraud transaction) $2,13,060.00
2 Average number of transactions per month detected as fraudulent by the model (TF) 28
3 Cost of providing customer executive support per fraudulent transaction detected by the model $1.50
4 Total cost of providing customer support per month for fraudulent transactions detected by the model (TF*$1.5) $42.00
5 Average number of transactions per month that are fraudulent but not detected by the model (FN) 203
6 Cost incurred due to fraudulent transactions left undetected by the model (FN*c) $1,07,590.00
7 Cost incurred per month after the model is built and deployed = (TF*$1.5) + [ (FN) * (avg amt / fraud transaction) ] $1,07,632.00
8 Final savings = Cost incurred before - Cost incurred after(1-7) $1,05,428.00

Credit Card Fraud Detection: Submitted by

Uploaded by

Copyright:

Available Formats

Credit Card Fraud Detection: Submitted by

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Credit Card Fraud Detection: Submitted by

Uploaded by

Copyright:

Available Formats

CREDIT

2. What do they do with it?

3. Where do the transactions happen?

4. When does it happen?

5. How is business affected by this?

Data Import Selecting

Validate model Model Fitting Compare model Validate model

ON TEST/UNSEEN • Assumption: Financial impact of 1% fraud in 1% of overall transaction

DATA • 1% of Fraudulent transactions = 0.01% overall transaction = 185 transactions.

Cost Benefit Analysis

Cost Benefit Analysis

You might also like

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.