Data Science
Build an AI-Powered Fraud Detection System for Banking Transactions
Simulate four weeks inside a bank's Fraud Risk & Intelligence team, building transaction baselines, investigating alert surges, diagnosing recurring fraud patterns, and presenting your findings to the Chief Risk Officer. 
Certified by
Role
Data Analyst
Industry
Education
No. of Subscribers
1
Level
Intermediate
Time Commitment
60 Hours
Duration
30 days
Tools you’ll learn
Here’s What You Work On
About the Company
You are stepping into the Fraud Risk & Intelligence team of a mid-sized retail bank. The bank processes millions of transactions daily across five customer segments - Retail Savings, Retail Current, SME Business, Premium Banking, and Digital-Only. Over the past quarter, three patterns have emerged that the team cannot fully explain.

You will work through them in sequence - building the baseline, investigating the alert surge, diagnosing the ATO recurrence, and finally synthesising everything into a fraud intelligence brief for the Chief Risk Officer.
Explore
the following work techniques
Fraud Data Analysis
Pattern Recognition & Hypothesis Testing
Model Explainability (SHAP)
Alert Investigation & SLA Tracking
Root Cause Analysis (5-Whys)
Bridging the gap
Over the past quarter, three patterns have emerged that the team cannot fully explain:

  • The Digital-Only segment has seen a sustained rise in fraud rate over 30 days, now approaching the critical threshold. The team does not yet know if this is a new fraud pattern establishing itself or a deterioration in detection logic.

  • A surge in fraud alerts during Q3 has revealed that 96% of Priority 1 alerts — the highest-value, fastest-response tier — are breaching their 5-minute acknowledgement target. The cost in delayed fraud containment is significant and growing.

  • Account Takeover fraud is recurring in the same customer segments despite repeated investigations. The team suspects device intelligence gaps and a specific overnight attack window, but has not yet proven it with data.

Apply
the following skills
Data Analysis
Data Visualization
Root cause analysis
Expected output
Across four components, you will produce ten portfolio deliverables that span the full fraud analyst workflow: 

  1. A colour-coded Fraud Baseline Report and written interpretation from Component 1; 
  2. An Alert Analysis Dashboard, SLA Breach Register, and written summary from Component 2; 
  3. A Feature Impact Summary, 5-Whys root cause writeup, and Recommendation Memo from Component 3; and 
  4. An Executive Summary, unified Fraud Intelligence Dashboard, and a 5-Slide Executive Briefing Deck for the Chief Risk Officer from Component 4. 

Together, demonstrating your ability to move from raw transaction data all the way through to a Board-ready fraud intelligence brief.
Create
the following deliverables
Fraud Baseline Report
Alert Analysis Dashboard
SLA Breach Register
Fraud Intelligence Dashboard
What you’ll need before starting
  • Basic proficiency in Excel or Python (pandas) — you should be comfortable loading a CSV, filtering rows, and computing averages. You do not need to be a programmer, but you must be able to work with tabular data without hand-holding.

  • Familiarity with at least one visualisation tool — Power BI, Tableau, or Python (Plotly/Matplotlib). You don't need to be an expert, but you should have built at least one chart before.