How nathprashant/credit_card_fraud_detection works
This repository documents a credit-card fraud detection workflow: obtain labeled transactions, explore and engineer features, train a Random Forest, evaluate it, and use a saved model to score new transactions. The tree contains the notebook and data/model directories, but not the documented src scripts or output directory. The workflow below reflects README-described behavior; detailed wiring and the prediction implementation cannot be verified from the sampled source.
Main entry points: main.ipynb Exploration · raw Raw CSV data · models Trained model
Components 12
Transaction data
- Raw CSV data
data/raw - Exploration
notebooks/main.ipynb - Feature engineering
notebooks/main.ipynb
Model workflow
- Random Forest training
notebooks/main.ipynb - Trained model
models - Model evaluation
notebooks/main.ipynb - Evaluation figures
Fraud scoring
- New transactions
- Transaction scoring
- Fraud predictions
Other components
- Analyst
- Kaggle dataset
Connections 15
- Analyst → Kaggle dataset: downloads dataset
- Analyst → Raw CSV data: places CSVs
- Kaggle dataset → Raw CSV data: provides transactions
- Analyst → Exploration: runs notebook
- Exploration → Raw CSV data: reads data
- Exploration → Feature engineering: prepares features
- Feature engineering → Random Forest training: provides features
- Random Forest training → Trained model: saves model
- Random Forest training → Model evaluation: evaluates classifier
- Model evaluation → Evaluation figures: saves figures
- Analyst → New transactions: provides transactions
- Analyst → Transaction scoring: runs scoring
- Transaction scoring → Trained model: loads model
- Transaction scoring → New transactions: scores rows
- Transaction scoring → Fraud predictions: returns predictions