💳 Online Payment Fraud Detection System

An intelligent Machine Learning project that detects fraudulent online payment transactions using multiple classification algorithms.

---

📖 Project Overview

Online payment fraud has become one of the biggest challenges in digital transactions. This project uses Machine Learning algorithms to classify whether a transaction is Legitimate or Fraudulent based on transaction details.

The system performs data preprocessing, feature encoding, feature scaling, model training, prediction, and performance evaluation using three popular Machine Learning algorithms.

---

🎯 Objectives

• Detect fraudulent online payment transactions.
• Compare multiple Machine Learning algorithms.
• Improve fraud detection accuracy.
• Reduce financial losses caused by fraudulent transactions.

---

🛠 Technologies Used

• Python
• Google Colab
• Pandas
• NumPy
• Matplotlib
• Seaborn
• Scikit-Learn

---

📂 Dataset Features

The dataset contains important transaction details such as:

• step – Represents the time step (hour) of the transaction.
• type – Type of transaction (e.g., CASH_IN, CASH_OUT, DEBIT, PAYMENT, TRANSFER).
• amount – Amount of money involved in the transaction.
• nameOrig – Unique identifier of the sender's account.
• oldbalanceOrg – Sender's account balance before the transaction.
• newbalanceOrig – Sender's account balance after the transaction.
• nameDest – Unique identifier of the receiver's account.
• oldbalanceDest – Receiver's account balance before the transaction.
• newbalanceDest – Receiver's account balance after the transaction.
• isFraud – Indicates whether the transaction is fraudulent (1) or legitimate (0).
• isFlaggedFraud – Indicates whether the transaction was flagged as suspicious by the system.

🎯 Target Variable

• isFraud
  • 0 → Legitimate Transaction
  • 1 → Fraudulent Transaction

---

⚙️ Machine Learning Algorithms Used

1. Logistic Regression
2. Decision Tree Classifier
3. Random Forest Classifier

---

📊 Model Performance

Machine Learning Model	Accuracy
Logistic Regression	78.33%
Decision Tree Classifier	93.67%
Random Forest Classifier	96.50% ✅

🏆 Best Performing Model

Random Forest Classifier

Accuracy: 96.50%

---

🔄 Project Workflow

Dataset
   │
   ▼
Data Preprocessing
   │
   ▼
Handling Missing Values
   │
   ▼
Encoding Categorical Data
   │
   ▼
Feature Scaling
   │
   ▼
Train-Test Split
   │
   ▼
Model Training
   │
   ▼
Prediction
   │
   ▼
Model Evaluation

---

📈 Evaluation Metrics

The models are evaluated using:

• Accuracy Score
• Confusion Matrix
• Classification Report

---

💳 Sample Predictions

The model predicts whether an online payment transaction is Fraudulent or Legitimate using two different transaction examples.

🧪 Example Prediction 1 (Fraudulent Transaction)

Input

Step               : 600
Transaction Type   : TRANSFER
Amount             : 950000
Old Balance Origin : 950000
New Balance Origin : 0
Old Balance Dest   : 0
New Balance Dest   : 950000
Flagged Fraud      : 0

Expected Output

Prediction : 1

Transaction Status : Fraudulent Transaction

A high-value transfer that empties the sender's account is predicted as a Fraudulent Transaction.

---

🧪 Example Prediction 2 (Legitimate Transaction)

Input

Step               : 250
Transaction Type   : PAYMENT
Amount             : 4500.75
Old Balance Origin : 15000.00
New Balance Origin : 10500.25
Old Balance Dest   : 5000.00
New Balance Dest   : 9500.75
Flagged Fraud      : 0

Expected Output

Prediction : 0

Transaction Status : Legitimate Transaction

A normal payment transaction with a reasonable amount and valid balance update is predicted as a Legitimate Transaction.

---

📁 Project Structure

Project_3_Online_Payment_Fraud_Detection_System_Using_Machine_Learning/
│
├── Online_Payment_Fraud_Detection_System.ipynb
├── Online_Payment_Fraud_Detection_System.py
├── Dataset/
│   ├── Online_Payment_Fraud_Dataset.csv
├── images/
│   ├── confusion_matrix.png
│   ├── feature_importance.png
│   └── model_accuracy_comparison.png
├── best_model.pkl
├── feature_names.pkl 
├── label_encoders.pkl
├── scaler.pkl 
├── requirements.txt
├── README.md
└── LICENSE

---

🚀 Installation

Clone the repository

git clone https://github.com/Tuhin092005/Project_3_Online_Payment_Fraud_Detection_System_Using_Machine_Learning.git

Navigate to the project

cd Project_3_Online_Payment_Fraud_Detection_System_Using_Machine_Learning

Install the required libraries

pip install -r requirements.txt

▶️ Run the Project

Using Python

python Online_Payment_Fraud_Detection_System.py

Using Google Colab

Open

Online_Payment_Fraud_Detection_System.ipynb

Upload:

• Online_Payment_Fraud_Dataset.csv

Run all cells.

---

📌 Future Improvements

• Deep Learning based fraud detection
• Real-time fraud monitoring
• Web Application using Flask or Streamlit
• API Integration
• Explainable AI (XAI)

---

👨‍💻 Author

Tuhin Maji

B.Tech CSE (Artificial Intelligence & Machine Learning)

Meghnad Saha Institute of Technology (MSIT)

---

⭐ If you found this project useful

Please consider giving this repository a ⭐ on GitHub.

It motivates future development and improvements.

---

📜 License

This project is licensed under the MIT License.

Feel free to use, modify, and distribute this project for educational purposes.