Credit Card Fraud Detection System
This project detects fraudulent online payment transactions using Machine Learning techniques.
The objective is to identify suspicious transactions accurately while handling highly imbalanced data.
Dataset
Kaggle Credit Card Fraud Dataset
Model
- Random Forest Classifier
- SMOTE for class imbalance handling
Evaluation on Unseen Data
The dataset was split into training and testing sets using an 80:20 ratio.
The test set was kept completely unseen during training.
Due to compute and environment limitations, detailed evaluation metrics
(Precision, Recall, F1-score, ROC-AUC) will be added in future updates.
The current implementation focuses on correct pipeline design, imbalance handling,
and inference workflow.
Real-Time Prediction Simulation
The trained model can simulate real-time fraud detection by scoring individual
transactions as input, mimicking how the model would behave in a live environment.
System Design & Scalability
Note: The system design can be extended to real-time pipelines using Kafka or
Spark Streaming in production environments to handle high-volume transaction streams.
