Fraud Detection Model

This repository contains a trained machine learning model for detecting fraudulent financial transactions. The model is saved as a .pkl file for easy loading and integration into applications such as Streamlit dashboards, APIs, or other real-time systems.

---

Overview

The fraud detection model was trained using a synthetic dataset of financial transactions. The model predicts whether a transaction is fraudulent or legitimate based on several key features like transaction type, amount, and account balances.

---

Features Used

The following features were used to train the model:

• step: Time in hours since the start of the simulation.
• type: Type of transaction (e.g., CASH-IN, CASH-OUT, TRANSFER, PAYMENT, DEBIT).
• amount: Amount of the transaction in local currency.
• oldbalanceOrg: Initial balance of the sender before the transaction.
• newbalanceOrig: New balance of the sender after the transaction.
• oldbalanceDest: Initial balance of the recipient before the transaction.
• newbalanceDest: New balance of the recipient after the transaction.
• isFlaggedFraud: Whether the transaction was flagged as suspicious based on business rules.

---

Model Details

• Model Type: XGBoost
• Library: Scikit-learn
• Performance:
  • Precision: 99.33%
  • Recall: 99.61%
  • F1 Score: 99.47%

Run the App Online (No Installation Needed)

Click the link below to access the app instantly on Streamlit Cloud: https://frauddetection-gwewmettkfc6ynvh2q4uzz.streamlit.app