scorecard-package 1.0.0

Python framework for credit risk scorecard development, validation, deployment and monitoring.

Scorecard Package

Python 3.9+

Version 1.0.0

MIT License

PyPI

Enterprise Credit Risk Scorecard Development Framework

Develop production-ready credit risk scorecards using industry-standard methodologies including Fine Classing (automatic fine binning) and Coarse Classing (user-controlled bin merging), WOE, IV, CSI, Correlation Analysis, VIF, Stepwise Logistic Regression, Logistic Regression, Weighted Logistic Regression, KS, GINI, Score Calibration, PSI, Validation and Deployment.

A comprehensive Python package for developing, validating, deploying and monitoring credit risk scorecards using industry-standard methodologies such as WOE, IV, KS, Logistic Regression, Weighted Logistic Regression, PSI and CSI.

---

Overview

Scorecard Package provides an end-to-end framework for building production-ready credit risk scorecards.

The package automates the complete scorecard development lifecycle including:

Data Preparation

• Fine Classing
• Coarse Classing
• Rule Management

Feature Engineering

• WOE
• IV
• KS

Model Development

• Correlation Analysis
• VIF
• Stepwise Regression
• Logistic Regression
• Weighted Logistic Regression

Validation & Monitoring

• PSI
• CSI
• Deployment

Designed for:

• Banks
• NBFCs
• FinTech Companies
• Credit Risk Analysts
• Data Scientists
• Model Validation Teams

---

Features

✅ Fine Classing

✅ Coarse Classing

✅ Automatic WOE Transformation

✅ Information Value (IV)

✅ KS Calculation

✅ Variable Bivariate Reports

✅ Correlation Analysis

✅ VIF Filtering

✅ Stepwise Regression

✅ Logistic Regression

✅ Weighted Logistic Regression

✅ Score Calibration

✅ Scorecard Generation

✅ Automatic Score Banding

✅ Model Validation

✅ Deployment Framework

✅ PSI Monitoring

✅ CSI Monitoring

✅ Export to Excel / CSV

---

Installation

pip install scorecard-package

Import the package:

import Scorecard_Package as sp

or install directly from GitHub

pip install git+https://github.com/sanket-shrishrimal/Scorecard-Package.git

---

Requirements

• Python 3.9+
• NumPy
• Pandas
• SciPy
• Statsmodels
• Scikit-learn
• OpenPyXL
• Matplotlib
• Joblib

---

Package Structure

Scorecard_Package
│
├── Fine_Classing.py
├── Coarse_Classing.py
├── Binning_Rules.py
├── Modelling.py
├── Validation.py
└── CSI.py

---

Workflow

Raw Data
     │
     ▼
Fine Classing
     │
     ▼
Coarse Classing
     │
     ▼
WOE Transformation
     │
     ▼
Correlation Analysis
     │
     ▼
VIF Filtering
     │
     ▼
Stepwise Selection
     │
     ▼
Logistic Regression
     │
     ▼
Score Calibration
     │
     ▼
Validation
     │
     ▼
Deployment
     │
     ▼
PSI / CSI Monitoring

---

Quick Start

import Scorecard_Package as sp

Fine Classing

import Scorecard_Package as sp

summary, bivariates, rules = sp.fine_classing(
    df=train_df,
    target="Default",
    ...
)

Coarse Classing

summary,bivariates,rules=sp.coarse_classing(...)

Model Development

results=sp.scorecard_model_pipeline(...)

Validation

results=sp.deploy_model(...)

CSI

csi_report=sp.calculate_csi(...)

---

Main Modules

Fine_Classing

Automatic fine bin creation with:

• IV
• KS
• WOE
• Bivariate Reports
• Fine Rules

---

Coarse_Classing

Automatic and manual bin merging.

Outputs:

• Coarse Rules
• WOE
• IV
• KS
• Monotonicity

---

Modelling

Complete scorecard development pipeline.

Includes:

• Correlation Analysis
• VIF
• Stepwise Selection
• Logistic Regression
• Weighted Logistic Regression
• Score Calibration
• Scorecard Generation

---

Validation

Production scoring module.

Automatically performs:

• Rule Loading
• WOE Transformation
• PD Prediction
• Score Generation
• Band Assignment
• PSI Calculation

---

CSI

Characteristic Stability Index calculation.

Supports:

• Development vs Validation comparison
• Stability Reports
• Excel Export

---

Outputs

The package automatically generates:

Module	Output
Fine Classing	Summary, Bivariates, Rules
Coarse Classing	Summary, Rules
Modelling	Model, Scorecard, Metrics
Validation	Validation Report
CSI	CSI Report

---

Why Scorecard Package?

Unlike generic machine learning libraries, Scorecard Package provides an end-to-end implementation of the complete credit scorecard lifecycle used across banks, NBFCs and financial institutions. The framework focuses on transparency, reproducibility and production-ready deployment rather than black-box modelling.

• End-to-end workflow
• Industry-standard methodology
• Transparent calculations
• Reusable rule engine
• Production deployment support
• Excel and CSV exports
• No hidden black-box modeling

---

Documentation

A comprehensive User Guide with detailed examples and case studies is currently under development.

---

Upcoming Features

• Reject Inferencing
• SHAP-based Model Explainability
• Model Monitoring Dashboard
• Interactive HTML Reports
• Automated Documentation
• Population Drift Reports
• Hyperparameter Optimization

---

Version

Current Version

1.0.0

---

Author

Sanket Shrishrimal

Credit Risk & Analytics Consultant

Mumbai, India

---

License

MIT License

---

Contributing

Contributions, feature requests and bug reports are welcome.

Please create an Issue or Pull Request.

---

Contact

For questions, feature requests or collaboration:

Email

[e-mail]

---

Acknowledgements

This package was developed to simplify and standardize enterprise credit scorecard development while remaining fully transparent and customizable for production use.