https://github.com/mehak-089/ai-powered-financial-fraud-detection-risk-analytics
Financial fraud is a multi-billion-dollar problem, requiring intelligent detection techniques. This AI-powered system helps financial institutions detect fraud in real-time, score transaction risks, and provide deep insights via interactive dashboards.
https://github.com/mehak-089/ai-powered-financial-fraud-detection-risk-analytics
finnancial fraud random-forest-classifier risk-analysis streamlit tableau tableau-public
Last synced: 3 months ago
JSON representation
Financial fraud is a multi-billion-dollar problem, requiring intelligent detection techniques. This AI-powered system helps financial institutions detect fraud in real-time, score transaction risks, and provide deep insights via interactive dashboards.
- Host: GitHub
- URL: https://github.com/mehak-089/ai-powered-financial-fraud-detection-risk-analytics
- Owner: Mehak-089
- Created: 2025-04-04T09:53:29.000Z (7 months ago)
- Default Branch: main
- Last Pushed: 2025-04-04T12:55:12.000Z (7 months ago)
- Last Synced: 2025-07-02T16:47:59.566Z (4 months ago)
- Topics: finnancial, fraud, random-forest-classifier, risk-analysis, streamlit, tableau, tableau-public
- Language: Jupyter Notebook
- Homepage:
- Size: 261 KB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# AI-Powered-Financial-Fraud-Detection-Risk-Analytics
🔍 **Technologies**: Python, Machine Learning, SQL, Tableau, Streamlit
🎯 **Ideal for**: Banking, FinTech, SaaS, E-Commerce Fraud Prevention
📈 **Project Type**: Resume-Ready, FAANG-Level, Data Science & ML
---
## 🎯 **Project Overview**
Financial fraud is a multi-billion-dollar problem, requiring intelligent detection techniques. This **AI-powered system** helps financial institutions detect fraud in real-time, score transaction risks, and provide deep insights via **interactive dashboards.**
💡 **Key Highlights:**
✅ **AI-based fraud detection** using Machine Learning
✅ **Risk scoring system** to rank high-risk transactions
✅ **Tableau Dashboard** for real-time analytics
✅ **SQL-based preprocessing** for efficiency
✅ **Streamlit Web App** for fraud prediction
---
## 📊 **Dataset**
📌 **Source**: [Kaggle - Credit Card Fraud Detection](https://www.kaggle.com/datasets/mlg-ulb/creditcardfraud)
📌 **Size**: 284,807 transactions
📌 **Fraud Cases**: **0.172%** (highly imbalanced dataset)
---
## 🔥 **Tech Stack & Tools**
| Stack | Tools Used |
|--------|-------------|
| **Programming** | Python, SQL |
| **Machine Learning** | Random Forest, Logistic Regression |
| **Libraries** | Pandas, NumPy, Scikit-learn, SHAP, Matplotlib |
| **Data Visualization** | Tableau |
| **Web App** | Streamlit |
| **Deployment** | Streamlit Cloud / Local |
---
## 🚀 **Project Workflow**
### 1️⃣ **Data Preprocessing & Feature Engineering**
🔹 Handled imbalanced dataset using undersampling
🔹 Normalized `Time` and `Amount` features
🔹 Identified key features using SHAP
### 2️⃣ **Machine Learning Model**
🔹 **Trained a Random Forest Classifier** for fraud detection
🔹 Computed **fraud probability & risk scores**
🔹 Saved trained model for deployment
### 3️⃣ **Streamlit Web App**
🔹 Upload transaction dataset
🔹 Predict fraud & risk scores
🔹 Generate model explanations using **SHAP**
### 4️⃣ **Tableau Dashboard**
📊 **Live Dashboard** → [Click Here](https://public.tableau.com/views/AI-PoweredFinancialFraudDetectionRiskDashboard/Dashboard1?:language=en-US&:sid=&:redirect=auth&:display_count=n&:origin=viz_share_link)
🔹 Fraud vs Non-Fraud Breakdown
🔹 Risk Heatmaps & High-Risk Users
🔹 Fraud Probability Distribution
---
## 📊 **Model Performance**
| Metric | Score |
|--------|------|
| 🎯 **Precision** | 94% |
| 🏆 **Recall** | 89% |
| 📈 **AUC-ROC** | 98% |
🔹 **Key Insights**:
✅ **High AUC-ROC (98%)** ensures strong fraud detection
✅ **Optimized false positives & false negatives**
✅ **Transaction Amount & Frequency** are top fraud indicators
---
## 📊 **Dashboard Insights**
🔗 **[Tableau Dashboard](https://public.tableau.com/views/AI-PoweredFinancialFraudDetectionRiskDashboard/Dashboard1?:language=en-US&:sid=&:redirect=auth&:display_count=n&:origin=viz_share_link)**
🔹 **Key Observations:**
📌 **Fraud Cases are Rare** → Only **94 transactions** flagged as fraud
📌 **High-Risk Transactions Have Distinct Scores** → Model effectively separates fraud vs legit
📌 **Fraud Probability Distribution is Skewed** → Most fraud cases have high probability
📌 **Transaction Amount & Risk Score Correlation** → Higher amounts often flagged as risky
---
## 🛠 **Setup & Installation**
### 1️⃣ Install Required Packages
```bash
pip install pandas scikit-learn shap streamlit
```
### 2️⃣ Run the Streamlit App
```bash
streamlit run streamlit_app/fraud_detector_app.py
```
### 3️⃣ Open the Tableau Dashboard
🔹 Load `fraud_predictions.csv` into **Tableau**
🔹 Apply filters & analyze fraud risk
---