🎯 Projects
Two end-to-end projects on real data. You deliver a reproducible notebook, a short report and a live presentation — plus the log of what you delegated to your code assistant and what you had to correct.
Overview
| Project 1 | Project 2 | |
|---|---|---|
| Weight | 40% | 60% |
| Scope | Supervised learning on a given business problem | End-to-end ML case on data you choose |
| Covers | Contents A, B, C | All contents, including D–G |
| Team | Groups of 2–3 | Same groups |
| Delivery | Notebook + report + presentation | Notebook + report + presentation |
| Released | Friday 9 October (Session 4) | Friday 30 October (Session 7) |
| Due | Thursday 22 October, 23:59 | Thursday 26 November, 23:59 |
| Presented | Friday 23 October, in class | Friday 27 November, in class |
.ipynb file) and the report by email to hizocar@gmail.com, with the subject Project X – Group name, before 23:59 on the due date.
📍 Project 1 — Prediction on a business problem
Project 1 · 40% of the final grade
Goal: take a business question, turn it into a supervised learning problem, and answer it with a model you can defend — including how you evaluated it and what it would mean to act on it.
Instructions
- Frame the problem. State the business question, the target variable, and what decision would change because of the prediction.
- Prepare the data. Load, clean and explore the dataset. Document every choice that removes or transforms rows.
- Build a baseline first. A trivial model (majority class, or the mean) — every later result is measured against it.
- Fit at least two models seen in Sessions 3–5, with a proper train/test split and cross-validation.
- Evaluate with the right metric for the business problem, and justify why it is the right one. Include a confusion matrix if it is a classification task.
- Interpret. Which variables drive the prediction, and does that make business sense?
- Write the assistant log (see Code Assistants).
Deliverables and deadline
- Reproducible notebook — runs top to bottom from a clean kernel.
- Report of max. 4 pages: problem, data, method, results, limitations.
- Assistant log (half a page).
- Presentation of 10 minutes + questions, in class on Friday 23 October.
Due: Thursday 22 October 2026, 23:59 · subject Project 1 – Group name.
Grading rubric
| Criterion | Weight |
|---|---|
| Problem framing: the question, the target and the decision are clear | 20% |
| Data preparation and exploration, with documented choices | 20% |
| Modelling: correct train/test discipline, sensible model choice, baseline | 25% |
| Evaluation: right metric, honest reading of the results | 20% |
| Communication: report, presentation, and answers about your own code | 15% |
📍 Project 2 — End-to-end machine learning case
Project 2 · 60% of the final grade
Goal: run a complete ML project on a dataset of your choice, from the business question to a recommendation, using the full toolbox of the course.
Instructions
- Choose the case and the data. Public datasets are fine (Kaggle, UCI, open government data); the business question must be your own. Propose it in class before you start.
- Explore and prepare the data, including the treatment of missing values and categorical variables.
- Apply at least three families of methods from the course — for example a linear or logistic model, a tree-based ensemble, and either shrinkage (LASSO/Ridge/Elastic Net) or an unsupervised step (PCA/clustering).
- Compare them honestly with cross-validation, and explain the trade-offs — not only which one scored highest, but which one you would deploy and why.
- Interpret and recommend. What should the business do differently because of this model? Quantify the expected effect, with explicit assumptions.
- State the limitations, including what would break in production and what data you would need next.
- Write the assistant log.
Deliverables and deadline
- Reproducible notebook — runs top to bottom from a clean kernel.
- Report of max. 8 pages.
- Assistant log (half a page).
- Presentation of 15 minutes + questions, in class on Friday 27 November.
Due: Thursday 26 November 2026, 23:59 · subject Project 2 – Group name.
Grading rubric
| Criterion | Weight |
|---|---|
| Relevance and framing of the chosen case | 15% |
| Data preparation and exploratory analysis | 15% |
| Breadth and correctness of the methods applied | 25% |
| Honest comparison and model selection | 20% |
| Business recommendation, quantified and with limitations | 15% |
| Communication and defence of your own code | 10% |
Reference
The Final Project notebook from previous years is available as a structural reference for the expected level of depth.
A model you cannot explain is not a finished project
Both presentations include questions about your own code and about the choices behind it. This is the core of the assessment: the assistant can write the pipeline, but only you can justify it.