marcellinus-witarsah/credit-score-modelling-with-optbinning

Credit Score Modelling: Perform a Weight of Evidence Logistic Regression Modelling (WoELR) to generate credit scorecard for loan approval with the help of optbinning library.

Credit Scorecard Modelling

Figure 1: Credit Score Illustration (Source).

Project Summary

In this project, we developed a credit score model leveraging Logistic Regression and Weight of Evidence techniques. The scoring methodology is based on the "point to double the odds" approach, utilizing Logistic Regression parameters, Weight of Evidence, and specific user-defined constraints to assign credit points for based on each predictor variable. Tools that will be used is optbinning which is a library for credit scorecard development.

Project Scope

The main objective is to create a reliable credit score model and develop a comprehensive credit scorecard.

Tools and Technologies

The project is built using Python 3.12.4, with the following libraries and tools:

1. pandas and numpy for data manipulation.
2. matplotlib and seaborn for data visualization.
3. optbinning for credit scorecard development (weight of evidence and information Value calculation).
4. scikit-learn, and optbinning for training and evaluation credit score model.

Installation and Setup

To run this project locally, you can use Anaconda. Ensure your Python version is 3.12.4. Recommended using linux environment for setting up environment. Then, install the required libraries from the requirements.txt file:

  make create_environment  # create conda environment
  conda activate credit-scorecard-modelling-with-optbinning  # access the environment
  make requirements  # install all libraries from the requirements.txt file
  make create_ipykernel  # create ipykernel

With this you can use run the code using the exact same dependencies that I used for this project.

For a detailed explanation of the project, please visit my Medium blog post.