⚡ Linear Regression with sklearn

These are Noah's notes on how to implement a linear regression model using the scikit-learn library. This demonstration will help you understand how scikit-learn simplifies the process of training, evaluating, and using linear regression models.

The interactive notebook includes:

• Synthetic data generated from a line we plant ourselves ($y = 4 + 3x$), so the result can be checked.
• Splitting the data into training and test sets with train_test_split.
• Training a LinearRegression model with a single call to fit(), and comparing what it learns with the planted line.
• Evaluation on unseen data with MSE and R², and a comparison with the training error.
• Visualization of the training data, the test data and the regression line.
• Predictions for new data, and the difference between interpolation and extrapolation.

Want to modify the code and experiment on your own?
Open the notebook in Google Colab

Bibliography

• scikit-learn API Reference – LinearRegression: Estimator parameters, attributes, and practical notes. https://scikit-learn.org/stable/modules/generated/sklearn.linear_model.LinearRegression.html
• scikit-learn User Guide – Train/Test Split: Best practices for train_test_split and validation strategies. https://scikit-learn.org/stable/modules/cross_validation.html#cross-validation
• scikit-learn Metrics – Regression: Definitions of the MSE and R² metrics used in the example. https://scikit-learn.org/stable/modules/model_evaluation.html#regression-metrics
• NumPy – Random sampling: np.random.rand, np.random.randn and np.random.seed, used to generate the synthetic data. https://numpy.org/doc/stable/reference/random/index.html