How hrishita456/limitforge works
LimitForge is a credit-limit decision pipeline: it loads customer history, engineers behavioral features, forecasts spend and estimates default risk, then combines those predictions into a guarded credit-limit recommendation. The main workflow is orchestrated by main.py; it also saves model artifacts, evaluation plots, and sample outputs. The README documents feature fan-out and the recommendation policy. Source excerpts show the orchestration and data-loader fallback; plot and artifact behavior is represented from the main.py calls to src/utils.py.
Main entry points: main.py Pipeline Runner · feature_engineering.py Feature Engineering · data_loader.py Dataset Loader · limit_recommendation_engine.py Limit Recommendation
Components 12
Customer Data
- Dataset Loader
src/data_loader.py - Synthetic Generator
src/synthetic_data.py - Feature Engineering
src/feature_engineering.py
Risk Modeling
- Spend Forecast
src/spend_forecast_model.py - Default Risk
src/default_risk_model.py
Recommendations and Outputs
- Pipeline Runner
main.py - Business Rule Target
src/limit_recommendation_engine.py - Limit Recommendation
src/limit_recommendation_engine.py - Artifact and Plot Helpers
src/utils.py
Other components
- Risk Team
- Customer History
- Recommendations and Evaluations
Connections 14
- Risk Team → Pipeline Runner: runs pipeline
README.md - Customer History → Dataset Loader: provides dataset
src/data_loader.py - Pipeline Runner → Dataset Loader: loads data
main.py - Dataset Loader → Synthetic Generator: generates fallback
src/data_loader.py - Pipeline Runner → Feature Engineering: engineers features
main.py - Feature Engineering → Spend Forecast: supplies features
README.md - Feature Engineering → Default Risk: supplies features
README.md - Pipeline Runner → Spend Forecast: trains and predicts
main.py - Pipeline Runner → Default Risk: trains and predicts
main.py - Pipeline Runner → Limit Recommendation: builds and trains
main.py - Limit Recommendation → Business Rule Target: builds target with
src/limit_recommendation_engine.py - Pipeline Runner → Limit Recommendation: requests recommendations
main.py - Pipeline Runner → Artifact and Plot Helpers: saves and plots
main.py - Pipeline Runner → Recommendations and Evaluations: writes outputs
main.py