Back to Projects
Credit Card Fraud & Risk Analysis
Analysed 10,000+ credit card transactions using Python to detect fraud patterns, building a model that hit 92% detection accuracy. Delivered a Power BI risk-scoring dashboard that cut manual review time by 40%.
10K+
Transactions Analyzed
92%
Detection Accuracy
40%
Review Time Cut
RFM
Segmentation Model
Problem Statement
Financial institutions need to understand customer behavior based on transaction data to optimize services and manage risk. This project aimed to:
- Identify common spending categories and patterns among cardholders
- Segment customers based on transaction behavior (high spenders, frequent travelers, specific merchant preferences)
- Uncover potential indicators of fraudulent activity or credit risk
- Visualize findings in an interactive dashboard for stakeholders
Solution & Key Insights
- Performed data wrangling and feature engineering using SQL to calculate metrics like transaction frequency, average transaction value, and spending by category
- Developed customer segments using RFM (Recency, Frequency, Monetary) analysis principles
- Identified correlations between demographic data and spending habits
- Analyzed transaction types and amounts to flag potential anomalies (unusually large purchases, rapid succession transactions)
- Built a fraud detection model in Python achieving 92% accuracy
- Power BI dashboard allowed filtering by time period, customer segment, spending category, and location
- Risk-scoring dashboard cut manual review time by 40%, helping the finance team catch issues weeks earlier
Conclusion & Impact
The analysis provided valuable insights into customer spending behavior. Marketing can use segment insights for targeted campaigns, risk management can monitor fraud indicators, and product development can identify popular spending categories. The 40% reduction in manual review time helped the finance team catch issues weeks earlier.