pushpak1203/credit_

How pushpak1203/credit_ works

Credit Sense is an offline, synthetic-data credit decisioning prototype for community lenders. Its core workflow generates applicant data, builds governed feature matrices, trains temporal-split models, and scores an application for approval, decline, or human review with borrower-facing reasons. Applicants can enter through a local dashboard or API; the API also records decisions and accepts appeals. Separate analysis stages evaluate model performance, fairness, drift, shadow-mode disagreement, and a five-stage pilot comparison. The README documents several analysis capabilities whose implementations are not excerpted; their links are shown only where index references support them.

Main entry points: features.py Feature registry · train.py Model training · main.py Scoring API · decision.py Decision engine

Components 17

Data and modeling

Synthetic data src/credit_sense/data/generate_synthetic.py
Feature registry src/credit_sense/features.py
Model training src/credit_sense/train.py

Scoring and explanations

Decision engine src/credit_sense/decision.py
Explanations src/credit_sense/explain.py
Reason codes config/reason_codes.yaml
Decision audit src/credit_sense/audit.py
Appeals workflow src/credit_sense/appeals.py
Appeals database src/credit_sense/appeals.py

Monitoring and pilot

Model evaluation src/credit_sense/evaluate.py
Fairness analysis src/credit_sense/fairness.py
Drift monitoring src/credit_sense/drift.py
Shadow comparison src/credit_sense/shadow.py
Pilot simulation src/credit_sense/pilot_simulation.py

Local interfaces

Scoring API api/main.py
Local dashboard app/dashboard.py

Other components

Applicant

Connections 24

Applicant → Scoring API: submits score api/main.py
Applicant → Local dashboard: enters application README.md
Synthetic data → Feature registry: provides records README.md
Feature registry → Model training: builds features src/credit_sense/train.py
Scoring API → Decision engine: calls scoring api/main.py
Local dashboard → Decision engine: calls scoring app/dashboard.py
Decision engine → Feature registry: builds features src/credit_sense/decision.py
Decision engine → Explanations: requests reasons src/credit_sense/decision.py
Explanations → Feature registry: builds features src/credit_sense/explain.py
Explanations → Reason codes: reads codes src/credit_sense/explain.py
Scoring API → Decision audit: appends decision api/main.py
Scoring API → Appeals workflow: submits appeal api/main.py
Local dashboard → Appeals workflow: lists appeals app/dashboard.py
Appeals workflow → Appeals database: reads and writes src/credit_sense/appeals.py
Feature registry → Model evaluation: builds features src/credit_sense/evaluate.py
Model training → Model evaluation: supplies models src/credit_sense/evaluate.py
Feature registry → Fairness analysis: builds features src/credit_sense/fairness.py
Model training → Fairness analysis: supplies models src/credit_sense/fairness.py
Feature registry → Drift monitoring: builds features src/credit_sense/drift.py
Feature registry → Shadow comparison: builds features src/credit_sense/shadow.py
Model training → Shadow comparison: supplies models src/credit_sense/shadow.py
Model evaluation → Pilot simulation: provides rule baseline src/credit_sense/pilot_simulation.py
Feature registry → Pilot simulation: builds features src/credit_sense/pilot_simulation.py
Model training → Pilot simulation: supplies models src/credit_sense/pilot_simulation.py

More Python diagrams: yujeong-star/rl_3dod · 5arth4k-x/redvect0r · 5ymph0en1x/sylc · ahnafnafee/songmirror · alniyatrui/cafe-ad

Drawn by GitDiagram on September 30, 2026 from the repository's file tree, README and source files. Read as Markdown ·