🤖 Machine learning service for predicting the risk of payment delinquency of a bank customer

The project was created as part of the "ML Service Development: From Idea to MVP" (RU) course, run by the team of online master's programme "Machine Learning and Data-Intensive Systems" of the Faculty of Computer Science of the Higher School of Economics.

Dataset | Model file | Streamlit web-application

📂 Files

• Credit_scoring.ipynb: the main Jupyter Notebook of the project, in which data analysis and model building were conducted
• streamlit_app/app.py: Streamlit application main file to run the web interface of the model
• streamlit_app/model.py: a script in which the model is loaded and the target variable is predicted

⚙️ Technologies

• Pandas and NumPy libraries were used for data processing.
• Data analysis and graphing were performed using the Matplotlib and Seaborn libraries.
• The scikit-learn library was used for machine learning, and in particular:
  • the RandomForestClassifier classification model
  • the RandomizedSearchCV method for finding the optimal hyperparameters of the model
  • methods for key model metrics estimating
  • the MinMaxScaler method for features scaling
• The pickle library was used to save the model.
• Using the Streamlit framework, a web service was created to interact with the model.

💻 Local run

To start the web interface, install the requirements and run the app.py file using the streamlit tool:

pip install -r requirements.txt
streamlit run ./streamlit_app/app.py

The application will then be available at http://localhost:8501/