How renteria-luis/fraud-detection-v1 works
Fraud Detection V1 scores PaySim transactions through a Streamlit demo or a FastAPI endpoint, which validates requests, loads the trained pipeline and operational threshold, and returns a fraud probability and decision. The retraining workflow loads and filters data, splits it, builds and fits a feature-engineering/model pipeline, then saves model and metadata artifacts. Evaluation metrics and the documented operational thresholding capability are represented separately. The API and training links follow sampled source; Streamlit internals and any connection from it to model artifacts were not shown, so that relationship is left unasserted.
Main entry points: train.py Training Workflow · main.py Scoring API · builder.py Pipeline Builder · config.py ML Configuration
Components 14
User Interfaces
- Streamlit Demo
app.py - Scoring API
api/main.py - Request Schemas
api/schemas.py
Data Preparation
- PaySim Loader
src/data/loader.py - Data Splitter
src/data/splitter.py - ML Configuration
src/config.py
Model Pipeline
- PaySim Features
src/features/engineering.py - Pipeline Builder
src/models/builder.py
Training and Evaluation
- Training Workflow
train.py - Model Evaluation
src/evaluation/metrics.py
Model Artifacts
- Trained Pipeline
models/fraud_detection_v1_xgb.pkl - Training Metadata
models/metadata_v1.json - Deployment Threshold
params.yaml
Other components
- Transaction User
Connections 15
- Transaction User → Streamlit Demo: uses demo
README.md - Transaction User → Scoring API: submits transaction
README.md - Scoring API → Request Schemas: validates and responds
api/main.py - Scoring API → Trained Pipeline: loads and scores
api/main.py - Scoring API → Deployment Threshold: loads threshold
api/main.py - Scoring API → Transaction User: returns prediction
api/main.py - Training Workflow → PaySim Loader: loads and filters
train.py - Training Workflow → Data Splitter: splits data
train.py - Training Workflow → Pipeline Builder: builds pipeline
train.py - Training Workflow → Trained Pipeline: saves pipeline
train.py - Training Workflow → Training Metadata: saves metrics
train.py - PaySim Loader → ML Configuration: reads settings
src/data/loader.py - Data Splitter → ML Configuration: reads settings
src/data/splitter.py - Pipeline Builder → PaySim Features: includes transformer
src/models/builder.py - Pipeline Builder → ML Configuration: uses feature lists
src/models/builder.py