Credit Scoring and Risk Prediction for Loan Optimization at Bati Bank Using Machine Learning techniques
This project aims to create a credit scoring model for Bati Bank, a leading financial service provider, in collaboration with a successful eCommerce platform. The goal is to enable a buy-now-pay-later service that allows customers to purchase products on credit if they qualify.
Bati Bank recognizes the growing demand for flexible payment solutions and aims to provide customers with the ability to buy products by credit. This requires developing a robust credit scoring model to assess the creditworthiness of potential borrowers. The model will utilize data from customer transactions to evaluate risk and determine suitable loan conditions.
- Define a proxy variable to categorize users as high risk (bad) or low risk (good).
- Select observable features that are strong predictors of the defined default variable.
- Develop a model that assigns risk probabilities for new customers.
- Develop a model that assigns credit scores based on risk probability estimates.
- Predict the optimal amount and duration of loans.
The data for this project is sourced from the Xente Challenge on Kaggle. The dataset includes the following fields:
TransactionId: Unique transaction identifier.BatchId: Unique batch number for processing.AccountId: Unique customer identifier on the platform.SubscriptionId: Unique identifier for customer subscriptions.CustomerId: Unique customer identifier.CurrencyCode: Currency of the transaction.CountryCode: Geographical code of the customer's country.ProviderId: Source provider of the purchased item.ProductId: Name of the purchased item.ProductCategory: Broader categories for products.ChannelId: Identifies the transaction channel (e.g., web, Android, iOS).Amount: Transaction value.Value: Absolute value of the transaction.TransactionStartTime: Timestamp of the transaction.PricingStrategy: Category of pricing structure.FraudResult: Fraud status (1 for yes, 0 for no).
src/: Contains source code for data loading, feature engineering, and EDA.tests/: Contains unit and integration tests.data/: Contains raw and processed data files..github/: Contains CI/CD configurations.notebook/: Contains the predction, preprocessing and analysis model.scripts/: Contains main script to run and visulazation.web-api/: Contains web-api fastapi and flask app for visulization.
1.Clone the repository:
git clone "https://github.com/Getachew0557/Credit-Score-Classifications.git"2.Navigate to the project directory:
cd Credit-Score-Classifications3.Install dependencies:
pip install -r requirements.txt4.Start the project:
python scripts/main.py5.To Forcast store sale use:
python flask-app/app.py-
Open your browser and past
http://127.0.0.1:5000 -
fill the form and submmit to the predction resul.
Step-by-step instructions to deploy to Render:
-
Sign up for a Render account and connect it to your GitHub repository.
-
Create a new web service by selecting your Credit-Score-Classifications repository.
-
Configure the service settings:
- select python language
- Set Start Command to:
python flask-app/app.py
-
Deploy the app and get the live URL.
-
Access the deployed app by visiting the Render URL https://credit-score-classifications-1.onrender.com/
For instructions on setting up and deploying the web API, see the Web API Deployment Guide.
Below is an example of the credit score and risk modeling predction visualization generated by the project:
Contributions are welcome! Please follow these steps:
- Fork the repository.
- Create a new branch (
git checkout -b feature-branch). - Make your changes and commit them (
git commit -m 'Add new feature'). - Push to the branch (
git push origin feature-branch). - Open a pull request.
- Name: Getachew Getu
- GitHub: Getachew0557
- Email: getachewgetu2010@gmail.com
- LinkedIn: Getachew Getu
This project is licensed under the Apache License 2.0. See the LICENSE file for more details.