Skip to content

Gowri-88/Transaction-Fraud-Risk-Intelligence-Analytics

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

8 Commits
 
 
 
 
 
 
 
 

Repository files navigation

Transaction Fraud Risk Intelligence Analytics

Overview

This project builds an end-to-end fraud risk analytics pipeline using the BankSim synthetic transaction dataset. The objective is to identify high-risk transactions through engineered behavioral signals and translate model output into business decision intelligence via a multi-page Power BI dashboard.

The workflow covers:

  • SQL-based data preparation
  • Feature engineering
  • Risk scoring framework
  • Performance evaluation metrics
  • Operational and financial impact visualization

This simulates a real financial risk analytics use case aligned with card network / digital payment monitoring environments.


Business Problem

Financial institutions process large transaction volumes where manual review capacity is limited. The challenge is to:

  • Detect fraudulent behavior early
  • Prioritize transactions for investigation
  • Minimize financial losses
  • Control operational review cost

This project builds a rule-driven risk scoring system to:

  • Segment transactions into actionable risk tiers
  • Evaluate detection performance
  • Quantify financial tradeoffs between intervention and loss

Dataset

  • Source: BankSim synthetic transaction dataset
  • Transaction-level behavioral data
  • Includes customer, merchant, amount, category, and fraud labels

Used for realistic fraud analytics experimentation without exposing sensitive financial data.


Architecture

CSV Dataset ↓ MySQL Ingestion (banksim_raw) ↓ Feature Engineering Tables ├── customer_velocity ├── customer_spend_baseline ├── category_risk └── zipcode_risk ↓ Risk Scoring Table └── transaction_risk_score ↓ Power BI Semantic Model ↓ Interactive Executive Dashboard


Risk Logic (Implemented)

This project uses interpretable rule-based scoring derived from behavioral signals.

1️⃣ Velocity Risk

Measures transaction frequency concentration.

  • tx_per_day ≥ 5 → Score 30
  • tx_per_day ≥ 3 → Score 20
  • Otherwise → Score 5

2️⃣ Amount Deviation Risk

Detects abnormal spending relative to user baseline.

  • amount ≥ 2× average → Score 40
  • amount ≥ 1.5× average → Score 25
  • Otherwise → Score 5

3️⃣ Category Risk

Captures systemic fraud exposure by transaction category.

  • fraud_rate ≥ 10% → Score 30
  • fraud_rate ≥ 5% → Score 20
  • Otherwise → Score 5

4️⃣ Final Risk Score

total_risk_score = velocity_risk

  • amount_risk
  • category_risk

5️⃣ Risk Tier Assignment

  • High → Score ≥ 85
  • Medium → Score ≥ 55
  • Low → Otherwise

This tier drives downstream prioritization and analytics.


Model Evaluation Metrics

Calculated using SQL aggregation:

  • Precision (High Risk)
  • Recall / Fraud Capture Rate
  • False Positive Rate
  • Tier-wise Fraud Rate

These metrics support assessment of detection effectiveness and operational impact.


Business Impact

This project demonstrates how raw transaction data can be converted into practical fraud decision support.

By implementing behavioral risk scoring and tier-based prioritization, organizations can:

  • Focus investigation efforts on the most economically risky transactions
  • Reduce manual review inefficiencies through structured risk segmentation
  • Quantify how much fraud is actually being captured
  • Understand the trade-off between detection coverage and false positives
  • Measure financial value generated from fraud prevention efforts

Instead of reacting to losses after they occur, this framework enables proactive risk management supported by measurable performance metrics and financial impact analysis. It establishes a scalable, explainable foundation for structured fraud monitoring and continuous performance optimization.

About

End-to-end fraud risk scoring and financial impact analytics using SQL feature engineering and Power BI modeling to prioritize investigation workflows and quantify loss prevention.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors