💳 Credit Card Fraud Detection Using Machine Learning

1. Project Overview

This project focuses on detecting fraudulent credit card transactions using machine learning. Fraudulent transactions are rare but financially damaging, making early detection crucial for banking security.
I use the Kaggle Credit Card Fraud dataset, which contains anonymized features (V1–V28, Time, Amount) and a target label Class (0 = legitimate, 1 = fraud).

The system provides two functionalities:

• Batch Prediction → Analyze multiple transactions at once
• Single Transaction Prediction → Interactively analyze individual transactions

2. Installation & Run

Clone repository

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

Install dependencies

pip install -r requirements.txt

Train the model

python train_model.py

Run the Streamlit app

streamlit run app.py

The application will be available in your browser at: Local URL: http://localhost:8501
Network URL: http://192.168.0.235:8501

Dataset

Download the dataset from Kaggle Credit Card Fraud Dataset
Place the Kaggle dataset (creditcard.csv) inside the data/ folder before running the project.

3. Dataset

• Source: Kaggle – Credit Card Fraud Detection
• Features:
  • V1–V28: anonymized PCA components
  • Time, Amount: transaction metadata
• Target: Class (0 = legitimate, 1 = fraud)
• Class Imbalance: ~99.8% legitimate vs. ~0.2% fraud

4. Data Preprocessing

• Scaling: Time and Amount scaled using StandardScaler
• Feature Order: Inputs arranged as [Time, V1–V28, Amount]
• Train-Test Split: 80/20 split for training and evaluation
• Handling Imbalance: Used ROC AUC and PR AUC instead of accuracy

5. Model Training

• Algorithm: RandomForestClassifier
• Why RandomForest? -> Robust to imbalance, handles non-linear patterns, outputs probability scores

Steps:

1. Scale Time and Amount
2. Train/test split
3. Train RandomForest
4. Evaluate with metrics
5. Save artifacts:
   • model.joblib → trained model
   • preprocess.joblib → scaler
   • metadata.json → threshold

Example Metrics:

• Accuracy: 1.00
• ROC AUC: 0.976
• PR AUC: 0.873
• F1-score (fraud class): 0.83

6. Streamlit Dashboard

The interactive Streamlit app (app.py) provides:

• Batch Prediction: Upload a CSV → model predicts and classifies transactions
• Single Prediction: Pre-filled random transaction → fraud probability + prediction (LEGIT or FRAUD)

7.Screenshots

Dashboard Screenshot

8. Conclusion

This project shows how machine learning can strengthen financial cybersecurity by effectively detecting fraudulent credit card transactions, even in highly imbalanced datasets. Using a RandomForest model with proper preprocessing and evaluation (ROC AUC, PR AUC), the system achieves strong fraud detection performance. From a security standpoint, it demonstrates how data-driven models can complement traditional rule-based systems, enhancing fraud prevention strategies. The interactive Streamlit app makes the solution practical for both technical and non-technical users, bridging the gap between machine learning research and real-world cybersecurity defense in financial systems.