TP: Creating a Linear Regression Package

A hands-on practical exercise on creating a simple machine learning package for linear regression, exploring package management, testing, manual publishing, and automating the publishing process.

Introduction

In this TP, you’ll create a simple machine learning package focused on linear regression. You’ll use modern Python tools like Poetry for package management, implement basic classes, write tests, set up continuous integration, manually publish your package, and finally automate the publishing process.

Objectives

By the end of this TP, you will:

1. Set up a Python project using Poetry
2. Implement a basic linear regression class
3. Write unit tests for your implementation
4. Use GitHub Actions for continuous integration
5. Manually publish your package to PyPI
6. Automate the publishing process using GitHub Actions

Prerequisites

• Basic understanding of Python and object-oriented programming
• Familiarity with linear regression concepts
• Git and GitHub account
• Python 3.8 or higher installed

Step 1: Setting Up the Project

Set up your project using Poetry.

1. Install Poetry if you haven’t already.
2. Create a new project named finance-ml.
3. Edit the pyproject.toml file to add necessary dependencies (numpy for calculations, pytest for testing).
4. Install the dependencies using Poetry.

Exercise 1: Set up the project structure as described above. Familiarize yourself with the pyproject.toml file and Poetry commands.

Step 2: Implementing Linear Regression

Implement a basic linear regression class.

Create a new file src/finance_ml/linear_models.py and implement a LinearRegression class with the following methods:

• __init__(self, use_intercept=True): Initialize the model parameters.
• fit(self, X, y): Fit the model to the training data.
• predict(self, X): Make predictions using the trained model.

Your implementation should handle cases with and without an intercept term and use numpy for efficient calculations.

Exercise 2: Implement the LinearRegression class as described above.

Step 3: Writing Tests

Write tests for your LinearRegression class to ensure it works correctly.

Create a new file tests/test_linear_models.py and write test functions to cover various scenarios and edge cases.

Exercise 3: Implement the test functions for your LinearRegression class.

Step 4: Setting Up Basic GitHub Actions

Set up a basic GitHub Actions workflow to run your tests automatically.

Create a new file .github/workflows/tests.yml:

name: Run Tests

on: [push, pull_request]

jobs:
  test:
    runs-on: ubuntu-latest
    steps:
    - uses: actions/checkout@v2
    - name: Set up Python
      uses: actions/setup-python@v2
      with:
        python-version: '3.8'
    - name: Install dependencies
      run: |
        pip install poetry
        poetry install
    - name: Run tests
      run: poetry run pytest

Exercise 4: 1. Set up a GitHub repository for your project and add the GitHub Actions workflow as shown above. 2. Push your code and verify that the tests run automatically on push and pull requests.

Step 5: Manual Package Publishing

Now, let’s publish your package to PyPI manually using Poetry.

1. Register an account on PyPI (https://pypi.org/)
2. Build your package using Poetry
3. Publish your package to PyPI using Poetry

Exercise 5: Follow the steps above to manually publish your package to PyPI. Verify that you can install your package from PyPI using pip.

Step 6: Automating Package Publishing

Finally, let’s automate the publishing process using GitHub Actions.

Exercise 6: 1. Modify the GitHub Actions workflow you created in Step 4 to include a job for publishing the package to PyPI when pushing to the main branch. 2. Research how to securely add your PyPI credentials as secrets in GitHub Actions. 3. Update your workflow to use these secrets for authentication when publishing. 4. Test your automated publishing process by pushing a change to the main branch.

Hint: You’ll need to add a new job to your workflow, use conditional execution based on the branch, and incorporate the PyPI credentials as secrets.