🚨 Fraud Detection AI System

An end-to-end machine learning system that detects fraudulent transactions using classification models, real-time prediction, analytics dashboard, and explainable AI.

This project demonstrates a production-style AI pipeline including preprocessing, model training, evaluation, feature importance, and batch prediction.

---

🧠 Features

• ✅ Real-time fraud prediction
• ✅ Fraud probability scoring
• ✅ Risk classification (Low / Medium / High)
• ✅ Batch fraud detection (CSV upload)
• ✅ Fraud analytics dashboard
• ✅ Feature importance visualization
• ✅ Model performance metrics
• ✅ Confusion matrix
• ✅ ROC curve
• ✅ Downloadable prediction results
• ✅ Explainable AI system

---

🏗️ System Architecture

Dataset → Preprocessing → Train Model → Save Model ↓ Prediction Engine ↓ Streamlit Dashboard

---

🛠️ Tech Stack

Machine Learning

• Scikit-learn
• Random Forest Classifier
• Feature Importance
• Model Evaluation Metrics

Data Processing

• Pandas
• NumPy
• StandardScaler
• OneHotEncoder

Visualization

• Matplotlib
• Streamlit Charts
• Analytics Dashboard

Deployment UI

• Streamlit

---

📂 Project Structure

fraud-detection-ai
│
├── app.py
├── train_model.py
├── preprocess.py
├── utils.py
├── requirements.txt
├── README.md
├── .gitignore
│
├── data/
│   └── fraud_data.csv
│
├── models/
│   ├── fraud_model.pkl
│   ├── feature_importance.pkl
│   └── model_metrics.pkl

---

🚀 How to Run Locally

1. Clone Repository

git clone https://github.com/imalkaperera13/fraud-detection-ai.git
cd fraud-detection-ai

2. Install Requirements

pip install -r requirements.txt

3. Train Model

python train_model.py

4. Run Application

streamlit run app.py

---

📊 Model Performance

The system evaluates the model using:

• Accuracy
• Precision
• Recall
• F1 Score
• Confusion Matrix
• ROC Curve

---

🔍 Example Prediction

Input:

Transaction Amount: 5000 Time: 02:00 AM Location: NewYork Device: Laptop

Output:

Fraud Probability: 0.92 Risk Level: HIGH Prediction: Fraud

---

📈 Dashboard Capabilities

• Fraud distribution chart
• Risk level analytics
• Transaction history
• Probability histogram
• Batch fraud detection results

---

🎯 Use Cases

• Banking fraud detection
• Credit card fraud detection
• Payment gateway monitoring
• E-commerce fraud detection
• Telecom fraud detection
• Insurance fraud detection

---

🧩 Machine Learning Pipeline

1. Load dataset
2. Data preprocessing
3. Feature encoding
4. Model training
5. Model evaluation
6. Feature importance extraction
7. Save trained model
8. Real-time prediction
9. Dashboard analytics

---

⭐ Why This Project Matters

This project demonstrates:

• End-to-end AI system design
• Machine learning pipeline
• Explainable AI
• Model evaluation
• Dashboard development
• Real-time prediction
• Industry-level fraud detection

---

🔮 Future Improvements

• XGBoost model integration
• Deep learning fraud detection
• Real-time API deployment
• Database integration
• User authentication
• Cloud deployment (AWS / Streamlit Cloud)

---

👨‍💻 Author

Imalka Perera

AI Engineer | DevOps | Machine Learning

GitHub: https://github.com/imalkaperera13