Modelpublic

amirsoahil101/Credit_Card_Fraud_Synthetic_Detector

sourceHugging Facemitupdated 27d agoView on Hugging Face

0likes

Model Card

💳 Credit Card Fraud & Anomaly Detection System

An interactive, production-ready Machine Learning Web Application built with Streamlit to detect fraudulent credit card transactions in real-time based on transformed PCA features and transaction metrics.

![Live App](https://credit-card-fraud-detection-system-md3f6bhxwpbrhz8ogcylwu.streamlit.app/)

🖼️ Application Preview & UI

[image]

Interactive Streamlit dashboard designed with business-friendly feature inputs, fault-tolerant logic, and instant fraud probability metrics.

🔗 Live Application

You can access and test the deployed application directly here: 👉 [Click Here to Launch Live Demo](https://credit-card-fraud-detection-system-md3f6bhxwpbrhz8ogcylwu.streamlit.app/)

🖼️ Key Features & Capabilities

—User-Friendly Interface: Business-oriented feature labels replacing raw mathematical PCA names ($V1$–$V5$) for seamless user experience.
—Smart Decision Dashboard: Clear status classification (Legitimate vs. Fraudulent), Risk Levels, and Model Confidence Score ($0-100\%$) instead of raw binary values ($0/1$).
—Fault-Tolerant Pipeline: Robust exception handling ensuring that the UI renders smoothly even if dependencies or .pkl files are missing.
—High Accuracy Benchmarks: Powered by high-precision ensemble tree classifiers trained on transaction risk patterns.

📊 Model Evaluation & Benchmarks

Multiple classification algorithms were evaluated to handle high variance and anomaly patterns. Here is the comparative accuracy metric across models:

Model Name	Accuracy Score	Status
Random Forest Classifier 🏆	0.980 (98.0%)	Selected Model
Extra Trees Classifier 🏆	0.980 (98.0%)	Top Performer
K-Neighbors Classifier	0.980 (98.0%)	Top Performer
XGBoost Classifier	0.980 (98.0%)	Top Performer
AdaBoost Classifier	0.975 (97.5%)	High Precision
Decision Tree Classifier	0.970 (97.0%)	Baseline
Gradient Boosting Classifier	0.970 (97.0%)	Baseline

Selected Baseline: The Random Forest Classifier achieved 98.0% accuracy with reliable probability estimation for risk grading.

⚙️ Model Features & Input Parameters

UI Parameter Name	Feature Code	Description / Domain Context
Transaction Timestamp	`Time`	Elapsed time in seconds since the first dataset transaction
Transaction Behavior Factor 1	`V1`	Primary PCA vector for transaction behavior patterns
Account Activity Vector 2	`V2`	PCA vector capturing account usage variations
Security Risk Index 3	`V3`	PCA vector measuring security variance
Anomalous Pattern Score 4	`V4`	PCA vector identifying spending behavior anomalies
Location/Device Metric 5	`V5`	PCA vector for contextual transaction parameters
Transaction Amount	`Amount`	Total monetary value of the transaction ($)
Target Output	`Class`	`0` = Legitimate Transaction, `1` = Fraudulent

📁 Repository Structure

text

├── credit_card_fraud_synthetic.csv     # Primary Transaction Dataset
├── model.ipynb                         # Data Preprocessing, EDA & Model Training
├── .gitignore                          # Git Ignore Rules
├── .gitattributes                      # Git LFS Configuration
├── app.py                              # Fault-Tolerant Streamlit Web Application
├── model.pkl                           # Trained Machine Learning Model
├── scaler.pkl                          # Fitted StandardScaler Object
├── column.pkl                          # Feature Column Definitions
├── requirements.txt                    # Python Dependencies
└── README.md                           # Project Documentation

bash

git clone https://github.com/amirsohail100/Credit-Card-Fraud-Detection-System.git

bash

cd Credit-Card-Fraud-Detection-System

bash

streamlit run app.py

bash

pip install -r requirements.txt

📄 License

This project is licensed under the MIT License.

📝 Author

👤 Amir Sohail

A high-accuracy Machine Learning project evaluating multiple classifiers (Random Forest, Extra Trees, XGBoost, KNN at 98% accuracy) on Credit Card Fraud Data. Features dynamic inputs like Time, PCA components (V1-V5), and Amount to predict transaction legitimacy (Class) via a fault-tolerant Streamlit web UI.