https://github.com/amanbig/fraud_detection

A comprehensive machine learning-based fraud detection system built with Streamlit, Scikit-learn, and XGBoost. This application provides real-time fraud detection capabilities with an intuitive web interface.
https://github.com/amanbig/fraud_detection

sckit-learn streamlit

Last synced: 3 months ago
JSON representation

A comprehensive machine learning-based fraud detection system built with Streamlit, Scikit-learn, and XGBoost. This application provides real-time fraud detection capabilities with an intuitive web interface.

• Host: GitHub
• URL: https://github.com/amanbig/fraud_detection
• Owner: Amanbig
• Created: 2025-09-27T11:03:48.000Z (9 months ago)
• Default Branch: main
• Last Pushed: 2025-09-27T11:10:11.000Z (9 months ago)
• Last Synced: 2026-03-30T02:44:33.722Z (3 months ago)
• Topics: sckit-learn, streamlit
• Language: Python
• Homepage:
• Size: 49.8 KB
• Stars: 0
• Watchers: 0
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

          
# 🔒 Fraud Detection System

A comprehensive machine learning-based fraud detection system built with **Streamlit**, **Scikit-learn**, and **XGBoost**. This application provides real-time fraud detection capabilities with an intuitive web interface.

## 📋 Table of Contents

- [Features](#features)

- [Dataset](#dataset)

- [Models](#models)

- [Installation](#installation)

- [Usage](#usage)

- [Project Structure](#project-structure)

- [Model Performance](#model-performance)

- [Screenshots](#screenshots)

- [Contributing](#contributing)

## ✨ Features

### 🔍 Single Transaction Prediction

- **Interactive Input Form**: Easy-to-use interface for entering transaction details

- **Dual Model Predictions**: Compare results from Logistic Regression and XGBoost

- **Real-time Results**: Instant fraud probability calculations

- **Visual Comparisons**: Bar charts showing model confidence levels

### 📊 Dataset Analytics

- **Comprehensive Statistics**: Key metrics including fraud rate and transaction patterns

- **Interactive Visualizations**: 

  - Fraud vs Legitimate transaction distribution

  - Transaction amounts by fraud status

  - Merchant category analysis

  - Location-based insights

### 📈 Model Performance Analysis

- **Accuracy Metrics**: Performance comparison between models

- **Probability Distributions**: Fraud score distributions for both models

- **Feature Importance**: XGBoost feature importance visualization

- **Confusion Matrices**: Model performance evaluation

## 📊 Dataset

The system uses a fraud detection dataset with the following features:

| Feature | Description | Type |

|---------|-------------|------|

| `TransactionID` | Unique transaction identifier | Numeric |

| `Amount` | Transaction amount in USD | Numeric |

| `Time` | Time since first transaction (seconds) | Numeric |

| `Location` | Transaction location | Categorical |

| `MerchantCategory` | Type of merchant | Categorical |

| `CardHolderAge` | Age of cardholder | Numeric |

| `IsFraud` | Target variable (0=Legitimate, 1=Fraud) | Binary |

### Data Preprocessing

- **Categorical Encoding**: OrdinalEncoder for location and merchant category

- **Missing Value Handling**: Mean imputation for numeric features

- **Feature Scaling**: StandardScaler for numeric features

- **Data Validation**: Robust handling of unknown categories

## 🤖 Models

### 1. Logistic Regression

- **Purpose**: Baseline model for interpretable predictions

- **Accuracy**: ~94.4%

- **Advantages**: Fast, interpretable, good baseline performance

- **Use Case**: Quick fraud screening and interpretable results

### 2. XGBoost Classifier

- **Purpose**: Advanced ensemble model for complex pattern recognition

- **Accuracy**: ~94.4%

- **Advantages**: Handles non-linear relationships, feature importance, robust performance

- **Use Case**: Production-ready fraud detection with high accuracy

## 🚀 Installation

### Prerequisites

- Python 3.8 or higher

- pip package manager

### Quick Start

1. **Clone or Download the Project**

   ```bash

   git clone https://github.com/Amanbig/fraud_detection

   cd assign_fraud

   ```

2. **Install Dependencies**

   ```bash

   pip install -r requirements.txt

   ```

3. **Prepare Models (First Time Only)**

   ```bash

   python retrain_models.py

   ```

4. **Run the Application**

   ```bash

   streamlit run fraud_detection_app.py

   ```

5. **Open Your Browser**

   - The app will automatically open at `http://localhost:8501`

   - If not, navigate to the URL shown in your terminal

### Manual Installation

```bash

pip install streamlit pandas numpy scikit-learn xgboost plotly seaborn matplotlib

```

## 📖 Usage

### 🔍 Making Predictions

1. **Navigate to Single Prediction Page**

2. **Enter Transaction Details**:

   - Transaction ID (numeric)

   - Amount in USD

   - Time (seconds since epoch)

   - Location (dropdown selection)

   - Merchant Category (dropdown selection)

   - Cardholder Age

3. **Click "Predict Fraud"**

4. **View Results**:

   - Both models' predictions

   - Fraud probability scores

   - Visual comparison chart

### 📊 Exploring Data

1. **Go to Dataset Overview Page**

2. **Review Key Metrics**:

   - Total transactions

   - Fraud cases count

   - Overall fraud rate

   - Average transaction amount

3. **Analyze Visualizations**:

   - Transaction distributions

   - Location patterns

   - Merchant category insights

### 📈 Model Analysis

1. **Visit Model Analytics Page**

2. **Compare Model Performance**:

   - Accuracy scores

   - Probability distributions

   - Feature importance (XGBoost)

   - Confusion matrices

## 📁 Project Structure

```

assign_fraud/

├── 📄 README.md                    # Project documentation

├── 🐍 fraud_detection_app.py       # Main Streamlit application

├── 🐍 retrain_models.py           # Model retraining script

├── 📊 model-deployment.ipynb      # Original development notebook

├── 📈 fraud-detection.csv         # Dataset file

├── 📋 requirements.txt            # Python dependencies

├── 🤖 logistic.pkl               # Trained Logistic Regression model

├── 🤖 xgboost.pkl                # Trained XGBoost model

├── ⚙️ ordinal_encoder.pkl        # Saved categorical encoder

├── ⚙️ scaler.pkl                 # Saved feature scaler

└── ⚙️ feature_names.pkl          # Saved feature names

```

## 📈 Model Performance

### Training Results

- **Dataset Size**: 500 transactions

- **Fraud Rate**: ~14% (realistic imbalanced dataset)

- **Train/Test Split**: 75/25

### Performance Metrics

| Model | Accuracy | Precision | Recall | F1-Score |

|-------|----------|-----------|--------|----------|

| Logistic Regression | 94.4% | High | Good | Good |

| XGBoost | 94.4% | High | Good | Good |

### Key Insights

- Both models achieve similar high accuracy

- XGBoost provides better feature importance insights

- Logistic Regression offers faster predictions and interpretability

- System handles class imbalance effectively

## 🖼️ Screenshots

### Main Dashboard

- Clean, professional interface

- Easy navigation between features

- Real-time prediction results

### Prediction Interface

- User-friendly input forms

- Dropdown selections for categorical features

- Clear fraud/legitimate indicators

### Analytics Dashboard

- Interactive charts and visualizations

- Model comparison tools

- Performance metrics display

## ⚙️ Configuration

### Model Retraining

To retrain models with your own data:

1. **Update Dataset**: Replace `fraud-detection.csv` with your data

2. **Run Retraining**: `python retrain_models.py`

3. **Restart App**: `streamlit run fraud_detection_app.py`

### Customization Options

- **Styling**: Modify CSS in the Streamlit app

- **Features**: Add/remove input fields as needed

- **Models**: Integrate additional ML algorithms

- **Visualizations**: Customize charts using Plotly

## 🔧 Troubleshooting

### Common Issues

1. **Model Loading Errors**

   ```bash

   python retrain_models.py  # Regenerate models

   ```

2. **Missing Dependencies**

   ```bash

   pip install -r requirements.txt --upgrade

   ```

3. **Dataset Not Found**

   - Ensure `fraud-detection.csv` is in the project directory

   - Check file encoding (use UTF-8)

4. **Port Already in Use**

   ```bash

   streamlit run fraud_detection_app.py --server.port 8502

   ```

## 🤝 Contributing

1. Fork the repository

2. Create a feature branch (`git checkout -b feature/AmazingFeature`)

3. Commit changes (`git commit -m 'Add AmazingFeature'`)

4. Push to branch (`git push origin feature/AmazingFeature`)

5. Open a Pull Request

## 📜 License

This project is licensed under the MIT License - see the [LICENSE](LICENSE) file for details.

## 🙏 Acknowledgments

- **Scikit-learn**: Machine learning library

- **XGBoost**: Gradient boosting framework

- **Streamlit**: Web app framework

- **Plotly**: Interactive visualizations

- **Seaborn & Matplotlib**: Statistical plotting

## 📞 Support

For questions or support, please:

1. Check the troubleshooting section

2. Review existing issues

3. Create a new issue with detailed description

---

**Built with ❤️ for fraud detection and financial security**

### Quick Commands Reference

```bash

# Install dependencies

pip install -r requirements.txt

# Retrain models (first time or after data changes)

python retrain_models.py

# Run the application

streamlit run fraud_detection_app.py

# Run on different port

streamlit run fraud_detection_app.py --server.port 8502

```

🔒 **Stay Safe, Detect Fraud!**