Day 01 — Part 2: Regression Algorithms¶
Regression is the foundation of predictive modelling. Before you can build a recommender, a churn predictor, or an anomaly detector, you need to understand how models learn to map inputs to a continuous output — and how to tell whether the resulting predictions are actually useful. This session covers the full regression toolkit: from the workhorse linear model to regularised variants to tree-based ensembles.
Session Overview
Difficulty: Beginner–Intermediate Reading time: ~2.5 hours | Exercises: ~2 hours Prerequisites: ML Basics (Day 01 Part 1) · NumPy array operations · Linear algebra basics (dot product, matrix multiplication) · Pandas DataFrames
Session Goals¶
By the end of this session you will be able to:
- Explain what regression is, what loss functions are, and why optimisation is at the heart of every regression model
- Fit a linear regression, read its coefficients, and diagnose whether its assumptions hold
- Choose between Ridge, Lasso, and ElasticNet based on the structure of your data — not just a formula
- Build and tune Decision Tree, Random Forest, and Gradient Boosting regressors
- Pick the right evaluation metric for a given problem and interpret residual plots honestly
Session Map¶
| # | Topic | File | Estimated Time |
|---|---|---|---|
| 1 | Regression Overview | 01-regression-overview.md |
15 min |
| 2 | Linear Regression | 02-linear-regression.md |
30 min |
| 3 | Ridge, Lasso, ElasticNet | 03-ridge-lasso-elasticnet.md |
30 min |
| 4 | Tree-Based Regression | 04-tree-based-regression.md |
30 min |
| 5 | Regression Metrics | 05-regression-metrics.md |
20 min |
| 6 | Exercises | 06-exercises.md |
45 min |
Total: ~2.5 hours including exercises
Prerequisites¶
Before this session you should be comfortable with:
- Python, NumPy, and Pandas basics (Week 01)
- Train/test splitting and the idea of generalisation error
- Basic matplotlib for plotting
Difficulty¶
Intermediate. The concepts build on one another — read in order. If you skip ahead to Random Forests without understanding linear regression's assumptions, you will not understand why tree models exist or when they are the better choice.