How ra-manuth/financial_fraud_detection-system works
This is a small fraud-classification application: a user enters transaction details in a Streamlit interface, requests a prediction, and sees a fraud/valid result. The README also describes an analysis and training workflow that targets TRANSFER and CASH_OUT transactions and produces a preprocessing-plus-logistic-regression pipeline. The notebook was not sampled, so its internal steps and the exact training-to-artifact wiring are documented rather than source-verified.
Main entry points: fraud_detection_model.pkl Trained pipeline · fraud_detection.py Streamlit interface · analysis_model.ipynb EDA and training
Components 10
Prediction app
- Streamlit interface
fraud_detection.py - Prediction request
fraud_detection.py - Prediction result
fraud_detection.py
Model pipeline
- Trained pipeline
fraud_detection_model.pkl - Feature preprocessing
fraud_detection_model.pkl - Logistic regression
fraud_detection_model.pkl
Analysis and training
- EDA and training
analysis_model.ipynb - High-risk filtering
analysis_model.ipynb
Other components
- Transaction user
- Kaggle transaction data
Connections 10
- Transaction user → Streamlit interface: enters details
fraud_detection.py - Streamlit interface → Trained pipeline: loads pipeline
fraud_detection.py - Streamlit interface → Prediction request: submits input
fraud_detection.py - Prediction request → Trained pipeline: predicts with
fraud_detection.py - Prediction request → Prediction result: displays prediction
fraud_detection.py - Kaggle transaction data → EDA and training: supplies data
Readme.md - EDA and training → High-risk filtering: analyzes transactions
Readme.md - EDA and training → Trained pipeline: trains pipeline
Readme.md - Trained pipeline → Feature preprocessing: includes
Readme.md - Trained pipeline → Logistic regression: includes
Readme.md